Introduction to Algorithms Exercises
I will put the actual implementation of exercises here.
I Foundations
1 The Role of Algorithms in Computing
1.1 Algorithms
1.11
Give a realworld example that requires sorting or a realworld example that requires computing a convex hull.
Skipped.
1.12
Other than speed, what other measures of efficiency might one use in a realworld setting?
Memory consumption.
1.13
Select a data structure that you have seen previously, and discuss its strengths and limitations.
Linked lists, where insertion and deletion take constant time, but locating an element by index takes linear time.
1.14
How are the shortestpath and travelingsalesman problems given above similar? How are they different?
They both need to minimize the total distance for travelling. But shortestpath is to find the shortest path between only two locations, while travelingsalesman is to find the shortest path containing all given locations.
1.15
Come up with a realworld problem in which only the best solution will do. Then come up with one in which a solution that is “approximately” the best is good enough.
Skipped.
1.2 Algorithms as a technology
1.21
Give an example of an application that requires algorithmic content at the application level, and discuss the function of the algorithms involved.
The Minesweeper game. You have to figure out if there is a mime under a button base on the numbers that you recovered. That is an algorithm.
1.22
Suppose we are comparing implementations of insertion sort and merge sort on the same machine. For inputs of size n, insertion sort runs in 8 \(n^2\) steps, while merge sort runs in 64 n lg n steps. For which values of n does insertion sort beat merge sort?
We can solve 8 \(n^2\) < 64 n lg n for n. Assume n ≥ 0, we get n < 8 lg n. With the help of Wolfram Alpha, we know that 1.1 < n < 43.5593. So for n ∈ [2, 43], insertion sort beats merge sort.
1.23
What is the smallest value of n such that an algorithm whose running time is 100 \(n^2\) runs faster than an algorithm whose running time is \(2^n\) on the same machine?
Solve 100 \(n^2\) < \(2^n\) for n (using Wolfram Alpha), we get 0.096704 < n < 0.103658 or n > 14.3247, so the smallest value of n is 0.
Question: Should I only consider positive integer values for n?
1.X Problems
11 Comparison of running times
For each function f(n) and time t in the following table, determine the largest size n of a problem that can be solved in time t, assuming that the algorithm to solve the problem takes f(n) microseconds.
1 second 1 minute 1 hour 1 day 1 month 1 year 1 century lg n \(\sqrt{n}\) n n lg n \(n^2\) \(n^3\) \(2^n\) n!
 1 second = \(10^6\) microseconds
 1 minute = 6 × \(10^7\) microseconds
 1 hour = 3.6 × \(10^9\) microseconds
 1 day = 8.64 × \(10^{10}\) microseconds
 1 month = 2.628 × \(10^{12}\) microseconds
 1 year = 3.154 × \(10^{13}\) microseconds
 1 century = 3.156 × \(10^{15}\) microseconds
1 second  1 minute  1 hour  1 day  1 month  1 year  1 century  

lg n  \(10^{301029.9957}\)  \(10^{18061799.7398}\)  \(10^{1083707984.3903}\)  \(10^{26008991625.368}\)  \(10^{791106828604.9426}\)  \(10^{9494486063241.967}\)  \(10^{950050666315524.8}\) 
\(\sqrt{n}\)  \(10^{12}\)  \(10^{15.5563}\)  \(10^{19.1126}\)  \(10^{21.873}\)  \(10^{24.8393}\)  \(10^{26.9977}\)  \(10^{30.9983}\) 
n  \(10^{6}\)  \(10^{7.7782}\)  \(10^{9.5563}\)  \(10^{10.9365}\)  \(10^{12.4196}\)  \(10^{13.4989}\)  \(10^{15.4991}\) 
n lg n  \(10^{4.7976}\)  \(10^{6.4474}\)  \(10^{8.1251}\)  \(10^{9.4401}\)  \(10^{10.8623}\)  \(10^{11.9019}\)  \(10^{13.8367}\) 
\(n^2\)  \(10^{3}\)  \(10^{3.8891}\)  \(10^{4.7782}\)  \(10^{5.4683}\)  \(10^{6.2098}\)  \(10^{6.7494}\)  \(10^{7.7496}\) 
\(n^3\)  \(10^{2}\)  \(10^{2.5927}\)  \(10^{3.1854}\)  \(10^{3.6455}\)  \(10^{4.1399}\)  \(10^{4.4996}\)  \(10^{5.1664}\) 
\(2^n\)  \(10^{1.2995}\)  \(10^{1.4123}\)  \(10^{1.5017}\)  \(10^{1.5603}\)  \(10^{1.6155}\)  \(10^{1.6517}\)  \(10^{1.7117}\) 
n!  \(10^{0.9636}\)  \(10^{1.0432}\)  \(10^{1.0984}\)  \(10^{1.1458}\)  \(10^{1.178}\)  \(10^{1.205}\)  \(10^{1.2421}\) 
The table is generated using following JavaScript code:
function generateTable() {
function binarySearch(f, target) {
const start = 0.000001;
const epsilon = 0.0000000001;
let left = start;
let right = start;
while (f(right) < target) {
right *= 2;
}
while (right  left > epsilon) {
const middle = left + (right  left) / 2;
const value = f(middle);
if (value < target) {
left = middle;
} else if (value > target) {
right = middle;
} else {
return middle;
}
}
return left + (right  left) / 2;
}
function normalize(x) {
return Math.round(x * 10000) / 10000;
}
const algorithms = [
{
label: "lg *n*",
func: (t) => Math.log10(2) * t
},
{
label: "\\\\(\\\\sqrt{n}\\\\)",
func: (t) => Math.log10(t) * 2
},
{
label: "*n*",
func: (t) => Math.log10(t)
},
{
label: "*n* lg *n*",
func: (t) => binarySearch(n => n * Math.pow(10, n) * Math.log2(10), t)
},
{
label: "\\\\(n^2\\\\)",
func: (t) => Math.log10(t) / 2
},
{
label: "\\\\(n^3\\\\)",
func: (t) => Math.log10(t) / 3
},
{
label: "\\\\(2^n\\\\)",
func: (t) => Math.log10(Math.log2(t))
},
{
label: "*n*!",
func: function (t) {
// TODO: use the Γ function (use Stirling's approximation?).
function fact(n) {
let result = 1;
let i = 2;
for (; i <= n; i++) {
result *= i;
}
if (i > n) {
result += result * (i  1) * (n + 1  i);
}
return result;
}
return binarySearch((x) => fact(Math.pow(10, x)), t);
}
}
];
const times = [
{
label: "1 second",
microseconds: 1e6
},
{
label: "1 minute",
microseconds: 6e7
},
{
label: "1 hour",
microseconds: 3.6e9
},
{
label: "1 day",
microseconds: 8.64e10
},
{
label: "1 month",
microseconds: 2.628e12
},
{
label: "1 year",
microseconds: 3.154e13
},
{
label: "1 century",
microseconds: 3.156e15
}
];
let result = `  ${times.map((x) => x.label).join("  ")} \n` +
`   ${times.map((x) => "").join("  ")} \n`;
for (const algorithm of algorithms) {
result += ` ${algorithm.label} `;
for (const time of times) {
result += ` \\\\(10^{${normalize(algorithm.func(time.microseconds))}}\\\\) `;
}
result += "\n";
}
return result;
}
2 Getting Started
2.1 Insertion sort
2.11
Using Figure 2.2 as a model, illustrate the operation of InsertionSort on the array A = ⟨31, 41, 59, 26, 41, 58⟩.
Skipped.
2.12
Rewrite the InsertionSort procedure to sort into nonincreasing instead of nondecreasing order.
Just change A[i] > key to A[i] < key in the original code.
2.13
Consider the searching problem:
Input: A sequence of n numbers A = ⟨\(a_1\), \(a_2\), …, \(a_n\)⟩ and a value v.
Output: An index i such that v = A[i] or the special value nil if v does not appear in A.
Write pseudocode for linear search, which scans through the sequence, looking for v. Using a loop invariant, prove that your algorithm is correct. Make sure that your loop invariant fulfills the three necessary properties.
LinearSearch(A, v)
 for i = 1 to A.length
 if A[i] == v
 return i
 return nil
Loop invariant: A[1‥i  1] does not contain value v. The only way the loop continues is that A[i] ≠ v, so we know A[1‥i] does not contain value v. Then we increase i by 1, so again, A[1‥i  1] still does not contain value v. If the loop is completed, i must be equal to A.length + 1, so the whole array does not contain value v, then we return nil.
If for some i, A[i] == v, we will find it in line 2 and return i in line 3. The only way to escape the loop is either for some i, A[i] == v, or none of the elements equals to v. we can guarantee that if there is an element in A, we will find it.
2.14
Consider the problem of adding two nbit binary integers, stored in two nelement arrays A and B. The sum of the two integers should be stored in binary form in an (n + 1)element array C. State the problem formally and write pseudocode for adding the two integers.
Problem: Array A and B only contain elements of 0 and 1, and A.length == B.length == n. Array C that have length n + 1. Rewrite the elements in C so that C only contains 0s and 1s, and \(∑_{i=1}^n A[i] × 2^{n  i} + ∑_{i=1}^n B[i] × 2^{n  i} = ∑_{i=1}^{n + 1} C[i] × 2^{n + 1  i}\).
Pseudocode:
AddBinary(A, B, C)
 carry = 0
 for i = 1 to A.length
 sum = A[n  i] + B[n  i] + carry
 C[n + 1  i] = sum mod 2
 carry = sum / 2
 C[0] = carry
2.2 Analyzing algorithms
2.21
Express the function \(n^3\) / 1000  100 \(n^2\)  100 n + 3 in terms of Θnotation.
Θ(\(n^3\)).
2.22
Consider sorting n numbers stored in array A by first finding the smallest element of A and exchanging it with the element in A[1]. Then find the second smallest element of A, and exchange it with A[2]. Continue in this manner for the first n  1 elements of A. Write pseudocode for this algorithm, which is known as selection sort. What loop invariant does this algorithm maintain? Why does it need to run for only the first n  1 elements, rather than for all n elements? Give the bestcase and worstcase running times of selection sort in Θnotation.
Solution is implemented here.
The loop invariant: at the start of each iteration of loop, The first i elements contains the smallest i elements in A, and they are in nondecreasing order.
It only need to run for first n  1 elements because after the loop, we have rearrange the smallest n  1 elements, to the front of A, so the last element must be the biggest one, so the whole array is ordered.
Bestcase and worstcase running times are both Θ(\(n^2\)).
2.23
Consider linear search again (see Exercise 2.13). How many elements of the input sequence need to be checked on the average, assuming that the element being searched for is equally likely to be any element in the array? How about in the worst case? What are the averagecase and worstcase running times of linear search in Θnotation? Justify your answers.
Assume the element to be searched is in the array, then the average elements to be checked is (n + 1) / 2.
Best case running time is Θ(1), worst case running time is Θ(n). If we are lucky, we can find the element at the first position, where only one element need to be checked. If we are unlucky, we can find the element at the last position, where all elements will be checked.
2.24
How can we modify almost any algorithm to have a good bestcase running time?
Precompute the result for some special cases, and test the inputs for such cases. If the input matches such cases we return the precomputed result.
2.3 Designing algorithms
2.31
Using Figure 2.4 as a model, illustrate the operation of merge sort on the array A = ⟨3, 41, 52, 26, 38, 57, 9, 49⟩.
Skipped.
2.32
Rewrite the Merge procedure so that it does not use sentinels, instead stopping once either array L or R has had all its elements copied back to A and then copying the remainder of the other array back into A.
See here for implementation.
2.33
Use mathematical induction to show that when n is an exact power of 2, the solution of the recurrence
T(n) = \(\begin{cases} 2 &\text{if $n = 2$,} \\ 2 T(n / 2) + n &\text{if $n = 2^k$, for $k > 1$} \end{cases}\)
is T(n) = n lg n.
 Base case: If n = 2, T(n) = 2. Since n lg n = 2 lg 2 = 2, T(n) = n lg n, so the claim holds.
 Inductive case: If n > 2, T(n) = 2 T(n / 2) + n, by induction, we know that T(n / 2) = (n / 2) lg (n / 2), so T(n) = 2 (n / 2) lg (n / 2) + n = n lg (n / 2) + n = n (lg n  1) + n = n lg n, The claim holds.
2.34
We can express insertion sort as a recursive procedure as follows. In order to sort A[1‥n], we recursively sort A[1‥n  1] and then insert A[n] into the sorted array A[1‥n  1]. Write a recurrence for the worstcase running time of this recursive version of insertion sort.
See here for implementation.
\[T(n) = \begin{cases} 1 &\text{if $n < 2$} \\ T(n  1) + Θ(n) &\text{if $n >= 2$)} \end{cases}\]
2.35
Referring back to the searching problem (see Exercise 2.13), observe that if the sequence A is sorted, we can check the midpoint of the sequence against v and eliminate half of the sequence from further consideration. The binary search algorithm repeats this procedure, halving the size of the remaining portion of the sequence each time. Write pseudocode, either iterative or recursive, for binary search. Argue that the worstcase running time of binary search is Θ(lg n).
See here for implementation.
BinarySearch(A, v)
 left = 1
 right = A.length + 1
 while left < right
 middle = ⌊(left + right) / 2⌋
 if A[middle] < v
 left = middle + 1
 else
 right = middle
 if left ≤ A.length and A[left] == v
 return left
 else
 return nil
After each iteration, the length of the searching range reduces by half, until the range is empty. So we have:
T(n) = \(c_1\), if n = 0;
T(n) = T(n / 2) + \(c_2\), if n > 0.
We prove T(n) = Θ(lg n) by induction:
 If n = 0, Θ(lg n) = Θ(lg 0) = Θ(∞), … Not sure how to go from here.
 If n > 0, Θ(lg n) = T(n / 2) + \(c_2\) = Θ(T(n / 2)). By induction, we know T(n / 2) = Θ(lg (n / 2)), so Θ(lg n) = Θ(lg (n / 2)) = Θ((lg n)  1) = Θ(lg n).
2.36
Observe that the while loop of lines 5–7 of the InsertionSort procedure in Section 2.1 uses a linear search to scan (backward) through the sorted subarray A[1‥j  1]. Can we use a binary search (see Exercise 2.35) instead to improve the overall worstcase running time of insertion sort to Θ(n lg n)?
No, we can not. Because despite the searching takes Θ(lg n) time, we still need to move n elements in the worstcase scenario, which takes Θ(n) time.
2.37 ★
Describe a Θ(n lg n)time algorithm that, given a set S of n integers and another integer x, determines whether or not there exist two elements in S whose sum is exactly x.
See here for implementations.
2.X Problems
21 Insertion sort on small arrays in merge sort
Although merge sort runs in Θ(n lg n) worstcase time and insertion sort runs in Θ(\(n^2\)) worstcase time, the constant factors in insertion sort can make it faster in practice for small problem sizes on many machines. Thus, it makes sense to coarsen the leaves of the recursion by using insertion sort within merge sort when subproblems become sufficiently small. Consider a modification to merge sort in which n / k sublists of length k are sorted using insertion sort and then merged using the standard merging mechanism, where k is a value to be determined.
 Show that insertion sort can sort the n / k sublists, each of length k, in Θ(n k) worstcase time.
 Show how to merge the sublists in Θ(n lg (n / k)) worstcase time.
 Given that the modified algorithm runs in Θ(n k + n lg (n / k)) worstcase time, what is the largest value of k as a function of n for which the modified algorithm has the same running time as standard merge sort, in terms of Θnotation?
 How should we choose k in practice?
 Sort a sublist of length k takes \(k^2\) time, so sorting n / k sublists takes (n / k) Θ(\(k^2\)) = Θ((n / k) \(k^2\)) = Θ(n k) time.
 Assume merging n sublists takes T(n) time, we have T(n) = 2 T(n) + \(c_1\) n, if n > 1. Also, T(n) = \(c_2\), if n = 1. Notice this is the same as equation 2.1 and 2.2. So we have T(n) = Θ(n lg n). So merging n / k sublists takes T(n / k) = Θ((n / k) lg (n / k)) = Θ(n lg (n / k)).
 We need to solve the equation n k + n lg (n / k) < c n lg (n). We can get k  lg k < (c  1) lg n from it. I think k < Θ(lg n), but I can’t prove it.
 With benchmarks and profiling.
22 Correctness of bubblesort
Bubblesort is a popular, but inefficient, sorting algorithm. It works by repeatedly swapping adjacent elements that are out of order.
Bubblesort(A)
 for i = 1 to A.length  1
 for j = A.length downto i + 1
 if A[j] < A[j  1]
 exchange A[j] with A[j  1]
Let A′ denote the output of Bubblesort(A). To prove that Bubblesort is correct, we need to prove that it terminates and that
A′[1] ≤ A′[2] ≤ … ≤ A′[n], (2.3)
where n = A.length. In order to show that Bubblesort actually sorts, what else do we need to prove?
The next two parts will prove inequality (2.3).
State precisely a loop invariant for the for loop in lines 2–4, and prove that this loop invariant holds. Your proof should use the structure of the loop invariant proof presented in this chapter.
Using the termination condition of the loop invariant proved in part (b), state a loop invariant for the for loop in lines 1–4 that will allow you to prove inequality (2.3). Your proof should use the structure of the loop invariant proof presented in this chapter.
What is the worstcase running time of bubblesort? How does it compare to the running time of insertion sort?

We also need to prove that the elements in A is the same as in A′.

The loop invariant: At the start of each iteration, A[j] is the smallest element in A[j‥A.length].
Proof:
 Initialization: Before the first iteration, j = A.length, So A[j] is the only element in A[j‥A.length], the claim holds.
 Maintenance:
 If A[j] < A[j  1], because we know that A[j] is the smallest element in A[j‥A.length], we can be sure A[j] is the smallest element in A[j  1‥A.length], after swapping A[j] and A[j  1], A[j  1] became the smallest element in A[j  1‥A.length]. After decreasing j, the loop invariant holds.
 If A[j] ≥ A[j  1], we know that A[j  1] is the smallest element in A[j  1‥A.length]. After decreasing j, the loop invariant holds.
 Termination: After termination, j = i, so we know that A[i] is the smallest element in A[i‥A.length].

The loop invariant: At the start of the loop, A[1‥i  1] is empty or contains the smallest i elements and are sorted.
Proof:
 Initialization: Before the first iteration, i = 1, So A[1‥i  1] is empty, the claim holds.
 Maintenance: After the inner loop, we know that A[i] is the smallest element in A[i‥A.length].
 If A[1‥i  1] is empty, i = 1, then A[1‥i] contains only one element and it is the smallest one in A[i‥A.length], so A[1‥i] is sorted and contains the smallest i element in A[1‥A.length]. After increasing i, the loop invariant holds.
 If A[1‥i  1] is not empty, then A[1‥i  1] contains the smallest i  1 element in A[1‥A.length] in sorted order, so A[i  1] ≤ A[i]. Because A[i] is the smallest element in A[i‥A.length], we know that A[1‥i] is sorted and contains the smallest i element in A[1‥A.length]. After increasing i, the loop invariant holds.
 Termination: After termination, i = A.length, and A[1‥A.length  1] contains the smallest A.length  1 elements in sorted order, so we know A[A.length  1] ≤ A[A.length], so the whole array is sorted.

Worstcase running time is Θ(\(n^2\)), it is the same as insertion sort. But insertion sort have a bestcase running time which is Θ(n), while the bestcase running time of bubble sort is still Θ(\(n^2\)).
23 Correctness of Horner’s rule
The following code fragment implements Horner’s rule for evaluating a polynomial
\(\begin{aligned} P(x) &= ∑_{k=0}^n a_k x^k \\ &=a_0 + x\left(a_1 + x\left(a_2 + … + x\left(a_{n  1} + x a_n\right) …\right)\right), \end{aligned}\)
given the coefficients \(a_0\), \(a_1\), …, \(a_n\) and a value for x:
 y = 0
 for i = n downto 0
 y = \(a_i\) + x ⋅ y
In terms of Θnotation, what is the running time of this code fragment for Horner’s rule?
Write pseudocode to implement the naive polynomialevaluation algorithm that computes each term of the polynomial from scratch. What is the running time of this algorithm? How does it compare to Horner’s rule?
Consider the following loop invariant:
At the start of each iteration of the for loop of lines 2–3,
y = \(\displaystyle ∑_{k = 0}^{n  (i + 1)} a_{k + i + 1} x^k\).
Interpret a summation with no terms as equaling 0. Following the structure of the loop invariant proof presented in this chapter, use this loop invariant to show that, at termination, y = \(∑_{k = 0}^n a_k x^k\).
Conclude by arguing that the given code fragment correctly evaluates a polynomial characterized by the coefficients \(a_0\), \(a_1\), …, \(a_n\).
See here for implementation.

Θ(n).

The pseudocode:
 y = 0
 for i = 0 to n
 p = \(a_i\)
 for j = 0 to i
 p = p ⋅ x
 y = y + p
The running time of this algorithm is Θ(\(n^2\)). It takes more time than Horner’s rule.

Proof:
 Initialization: Before the first iteration, i = n, y = \(∑_{k = 0}^{n  (i + 1)} a_{k + i + 1} x^k = ∑_{k = 0}^{1} a_{k + n + 1} x^k = 0\), so the claim holds.
 Maintenance: After line 3, \(y' = a_i + x ⋅ y = a_i + x \left(∑_{k = 0}^{n  (i + 1)} a_{k + i + 1} x^k\right) = a_i ⋅ x^0 + ∑_{k = 0}^{n  (i + 1)} a_{k + i + 1} x^{k + 1} = a_i ⋅ x^0 + ∑_{k = 1}^{n  i} a_{k + i} x^k = ∑_{k = 0}^{n  i} a_{k + i} x^k\). After decreasing i, the claim holds.
 Termination: At termination, i = 1, so y = \(∑_{k = 0}^{n  ((1) + 1)} a_{k + (1) + 1} x^k = ∑_{k = 0}^n a_k x^k\).

I thought I have proved it at step 3.
24 Inversions
Let A[1‥n] be an array of n distinct numbers. If i < j and A[i] > A[j], then the pair (i, j) is called an inversion of A.
 List the five inversions of the array ⟨2, 3, 8, 6, 1⟩.
 What array with elements from the set {1, 2, …, n} has the most inversions? How many does it have?
 What is the relationship between the running time of insertion sort and the number of inversions in the input array? Justify your answer.
 Give an algorithm that determines the number of inversions in any permutation on n elements in Θ(n lg n) worstcase time. (Hint: Modify merge sort.)

The five inversions are (1, 5), (2, 5), (3, 4), (3, 5) and (4, 5).

The array ⟨n, …, 2, 1⟩ has the most inversions. It has n × (n  1) / 2 inversions.

Let k be the inversion of an array, the the running time of insertion sort on it is Θ(k).
Let \(k_i\) be the numbers of inversions whose second element is i. The total sorting time is \(∑_{i = 1}^n\left(c_1 k_i + c_2\right) = c_1 k + c_2 n = Θ(k)\).

See here for implementation.
3 Growth of Functions
3.1 Asymptotic notation
Notation  Definition 

f(n) = O(g(n))  ∃ c > 0, \(n_0\) > 0: ∀ n ≥ \(n_0\): 0 ≤ f(n) ≤ c g(n) 
f(n) = Ω(g(n))  ∃ c > 0, \(n_0\) > 0: ∀ n ≥ \(n_0\): 0 ≤ c g(n) ≤ f(n) 
f(n) = Θ(g(n))  ∃ \(c_1\) > 0, \(c_2\) > 0, \(n_0\) > 0: ∀ n ≥ \(n_0\): 0 ≤ \(c_1\) g(n) ≤ f(n) ≤ \(c_2\) g(n) 
f(n) = o(g(n))  ∀ c > 0: ∃ \(n_0\) > 0: ∀ n ≥ \(n_0\): 0 ≤ f(n) < c g(n) 
f(n) = ω(g(n))  ∀ c > 0: ∃ \(n_0\) > 0: ∀ n ≥ \(n_0\): 0 ≤ c g(n) < f(n) 
3.11
Let f(n) and g(n) be asymptotically nonnegative functions. Using the basic definition of Θnotation, prove that max(f(n), g(n)) = Θ(f(n) + g(n)).
In the following statements, n is big enough that both f(n) and g(n) is nonnegative.
Because f(n) ≤ max(f(n), g(n)), and g(n) ≤ max(f(n), g(n)), we know that f(n) + g(n) ≤ 2 max(f(n), g(n)). So 0.5 (f(n) + g(n)) ≤ max(f(n), g(n)).
Because f(n) ≤ f(n) + g(n), and g(n) ≤ f(n) + g(n), we know that max(f(n), g(n)) ≤ f(n) + g(n).
So we have 0.5 (f(n) + g(n)) ≤ max(f(n), g(n)) ≤ f(n) + g(n), max(f(n), g(n)) = Θ(f(n) + g(n)).
3.12
Show that for any real constants a and b, where b > 0,
\((n + a)^b\) = Θ(\(n^b\)). (3.2)
We want to find constant \(c_1\), \(c_2\) and \(n_0\) so that if n > \(n_0\), \(c_1 n^b ≤ (n + a)^b ≤ c_2 n^b\).
\(c_1 n^b ≤ (n + a)^b ≤ c_2 n^b\)
⇔ \(\left({c_1}^{1 / b}\right)^b n^b ≤ (n + a)^b ≤ \left({c_2}^{1 / b}\right)^b n^b\)
⇔ \(\left({c_1}^{1 / b} n\right)^b ≤ (n + a)^b ≤ \left({c_2}^{1 / b} n\right)^b\)
⇔ \({c_1}^{1 / b} n ≤ n + a ≤ {c_2}^{1 / b} n\)
⇔ \({c_1}^{1 / b} n  n ≤ a ≤ {c_2}^{1 / b} n  n\)
⇔ \(\left({c_1}^{1 / b}  1\right) n ≤ a ≤ \left({c_2}^{1 / b}  1\right) n\)
We need n to be greater than some \(n_0\), so we should have \({c_1}^{1 / b}\)  1 < 0, and \({c_2}^{1 / b}\)  1 > 0, then we have n ≥ \(\frac{a}{{c_1}^{1 / b}  1}\), and n ≥ \(\frac{a}{{c_1}^{1 / b}  1}\), i.e. n ≥ max\(\left(\frac{a}{{c_1}^{1 / b}  1}, \frac{a}{{c_2}^{1 / b}  1}\right)\). Let \(c_1 = \left(\frac{1}{2}\right)^b\), \(c_2 = 2^b\), we have n ≥ max(2 a, a). So \(n_0\) can be max(2 a, a).
Formally, for any n > max(2 a, a), \(\left(\frac{1}{2}\right)^b n^b ≤ (n + a)^b ≤ 2^b n^b\), \(n + a)^b = Θ(n^b)\).
3.13
Explain why the statement, “The running time of algorithm A is at least O(\(n^2\)),” is meaningless.
It is like saying x is at least less than or equal to 10.
3.14
Is \(2^{n + 1} = O\left(2 ^ n\right)\)? Is \(2^{2 n} = O\left(2^n\right)\)?
\(2^{n + 1} = 2 × 2^n = O\left(2 ^ n\right)\), \(2^{2 n} = \left(2^n\right)^2 ≠ O\left(2^n\right)\).
3.15
Prove Theorem 3.1.
If f(n) = Θ(g(n)), c1 g(n) ≤ f(n) ≤ c2 g(n), for all n ≥ n0, for some c1, c2 and n0. Because f(n) ≤ c2 g(n), we know that f(n) = O(g(n)). Because c1 g(n) ≤ f(n), we know that f(n) = Ω(g(n)).
If f(n) = O(g(n)) then f(n) ≤ c2 g(n), for all n ≥ n0, for some c2 and n0. If f(n) = Ω(g(n)), then c1 g(n) ≤ f(n), for all n ≥ n1, for some c1 and n1. So c1 g(n) ≤ f(n) ≤ c2 g(n), for all n ≥ max(n0, n1).
3.16
Prove that the running time of an algorithm is Θ(g(n)) if and only if its worstcase running time is O(g(n)) and its bestcase running time is Ω(g(n)).
Running time of an algorithm is Θ(g(n)) means the running time is bounded by a function f(n) that c1 g(n) ≤ f(n) ≤ c2 g(n), for all n ≥ n0, for some c1, c2 and n0 > 0. So the worstcase running time is bounded by c2 g(n), and the bestcase running time is bounded by c1 g(n). So the worstcase running time is O(g(n)), and the bestcase running time is Ω(g(n)).
If the worstcase running time is O(g(n)), it means the running time is bounded by a function f1(n) from above that f1(n) ≤ c2 g(n) for sufficiently large n for some c2. If the bestcase running time is Ω(g(n)), it means the running time is bounded from below by a function f2(n) that c1 f(n) ≤ f2(n) for sufficiently large n for some c1. Because f1(n) and f2(n) is the worstcast running time and the bestcase running time, the running time is bounded by c1 g(n) c2 g(n), so the running time is Θ(g(n)).
3.17
Prove that o(g(n)) ∩ ω(g(n)) is the empty set.
Assume there exist a function f(n) that f(n) = o(g(n)) and f(n) = ω(g(n)), we have:
 For all c1 > 0, for some n0 > 0, for all n ≥ n0, c1 f(n) < g(n).
 For all c2 > 0, for some n1 > 0, for all n ≥ n1, c2 f(n) > g(n).
Let c2 = c1, n = max(n0, n1), we have c1 f(n) < g(n) and c1 f(n) > g(n) which is impossible, so f(n) does not exist. So o(g(n)) ∩ ω(g(n)) is the empty set.
3.18
We can extend our notation to the case of two parameters n and m that can go to infinity independently at different rates. For a given function g(n, m), we denote by O(g(n, m)) the set of functions
O(g(n, m)) = {f(n, m) : there exist positive constants c, \(n_0\), and \(m_0\) such that 0 ≤ f(n, m) ≤ c g(n, m) for all n ≥ \(n_0\) or m ≥ \(m_0\)}.
Give corresponding definitions for Ω(g(n, m)) and Θ(g(n, m)).
Ω(g(n, m)) = {f(n, m) : there exist positive constants c, \(n_0\), and \(m_0\) such that 0 ≤ c g(n, m) ≤ f(n, m) for all n ≥ \(n_0\) or m ≥ \(m_0\)}.
Θ(g(n, m)) = {f(n, m) : there exist positive constants \(c_1\), \(c_2\), \(n_0\), and \(m_0\) such that 0 ≤ \(c_1\) g(n, m) ≤ f(n, m) ≤ \(c_2\) g(n, m) for all n ≥ \(n_0\) or m ≥ \(m_0\)}.
3.2 Standard notations and common functions
3.21
Show that if f(n) and g(n) are monotonically increasing functions, then so are the functions f(n) + g(n) and f(g(n)), and if f(n) and g(n) are in addition nonnegative, then f(n) ⋅ g(n) is monotonically increasing.
 m ≤ n ⇒ (f(m) ≤ f(n)) ∧ (g(m) ≤ g(n)) ⇒ f(m) + g(m) ≤ f(n) + g(n).
 m ≤ n ⇒ g(m) ≤ g(n) ⇒ f(g(m)) ≤ f(g(n)).
 (m ≤ n) ∧ (∀ x: f(x) ≥ 0) ∧ (∀ x: g(x) ≥ 0)
⇒ (f(m) ≤ f(n)) ∧ (g(m) ≤ g(n)) ∧ (∀ x: f(x) ≥ 0) ∧ (∀ x: g(x) ≥ 0)
⇒ f(m) ⋅ g(m) ≤ f(n) ⋅ g(m) ≤ f(n) ⋅ g(n)
⇒ f(m) ⋅ g(m) ≤ f(n) ⋅ g(n).
3.22
Prove equation (3.16).
\(a^{\log_b c} = \left(c^{\log_c a}\right)^{\log_b c} = c^{\left(\log_c a\right)\left(\log_b c\right)} = c^{\frac{\ln a}{\ln c} \frac{\ln c}{\ln b}} = c^{\frac{\ln a}{\ln b}} = c^{\log_b a}\).
3.23
Prove equation (3.19). Also prove that n! = ω(\(2^n\)) and n! = o(\(n^n\)).
Proving equation (3.19):
According to equation (3.18), we know that
\(\sqrt{2 \pi n} \left(\frac{n}{e}\right)^n \left(1 + \frac{c_1}{n}\right) ≤ n! ≤ \sqrt{2 \pi n} \left(\frac{n}{e}\right)^n \left(1 + \frac{c_2}{n}\right)\)
So
\(\ln \left(\sqrt{2 \pi n} \left(\frac{n}{e}\right)^n \left(1 + \frac{c_1}{n}\right)\right)
≤ \ln (n!)
≤ \ln \left(\sqrt{2 \pi n} \left(\frac{n}{e}\right)^n \left(1 + \frac{c_2}{n}\right)\right)\)
⇒ \(\ln \left(\left(\frac{n}{e}\right)^n\right) + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_1}{n}\right)\right)
≤ \ln (n!)
≤ \ln \left(\left(\frac{n}{e}\right)^n\right) + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_2}{n}\right)\right)\)
⇒ \(n \ln \left(\frac{n}{e}\right) + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_1}{n}\right)\right)
≤ \ln (n!)
≤ n \ln \left(\frac{n}{e}\right) + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_2}{n}\right)\right)\)
⇒ \(n \left(\ln n  1\right) + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_1}{n}\right)\right)
≤ \ln (n!)
≤ n \left(\ln n  1\right) + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_2}{n}\right)\right)\)
⇒ \(n \ln n  n + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_1}{n}\right)\right)
≤ \ln (n!)
≤ n \ln n  n + \ln \left(\sqrt{2 \pi n} \left(1 + \frac{c_2}{n}\right)\right)\)
⇒ \(n \ln n  n + \frac{1}{2} \ln {(2 \pi n)} + \ln {\left(1 + \frac{c_1}{n}\right)}
≤ \ln (n!)
≤ n \ln n  n + \frac{1}{2} \ln {(2 \pi n)} + \ln {\left(1 + \frac{c_2}{n}\right)}\).
So we have ln(n!) = Ω(n ln n), and ln(n!) = O(n ln n), so ln(n!) = Θ(n ln n).
Proving n! = ω(\(2^n\)):
\(\lim_{n → ∞}\frac{2^n}{n!} = \lim_{n → ∞}\frac{2 × 2 × 2 × 2 × 2 × … × 2}{1 × 2 × 3 × 4 × 5 × … × n} = \lim_{n → ∞}\left(\frac{2}{1} × \frac{2}{2} × \frac{2}{3} × \frac{2}{4} × \frac{2}{5} × … × \frac{2}{n}\right) = 2 \lim_{n → ∞}\left(\frac{2}{3} × \frac{2}{4} × \frac{2}{5} × … × \frac{2}{n}\right) ≤ 2 \lim_{n → ∞}\left(\frac{2}{n}\right) = 0\).
Proving n! = o(\(n^n\)):
\(\lim_{n → ∞}\frac{n!}{n^n} = \lim_{n → ∞}\frac{1 × 2 × 3 × 4 × 5 × … × n}{n × n × n × n × n × … × n} = \lim_{n → ∞}\left(\frac{1}{n} × \frac{2}{n} × \frac{3}{n} × \frac{4}{n} × \frac{5}{n} × … × \frac{n}{n}\right) ≤ \lim_{n → ∞}\left(\frac{1}{n}\right) = 0\).
3.24 ★
Is the function ⌈lg n⌉! polynomially bounded? Is the function ⌈lg lg n⌉! polynomially bounded?
Skipped.
3.25 ★
Which is asymptotically larger: lg(\(\lg^* n\)) or \(\lg^* \left(\lg n\right)\)?
By the definition of \(\lg^*\), we have
\(\lg^* n = \begin{cases}0&n ≤ 1\\ \lg^* \left(\lg n\right) + 1&n > 1\end{cases}\).
So \(\lg^* n = Θ\left(\lg^* \left(\lg n\right)\right)\). Because \(\lg^* n\) is asymptotically larger than \(\lg \left(\lg^* n\right)\), we know that \(\lg^* \left(\lg n\right)\) is asymptotically larger than \(\lg \left(\lg^* n\right)\).
3.26
Show that the golden ratio ϕ and its conjugate \(\hat{\phi}\) both satisfy the equation \(x^2 = x + 1\).
\(\phi^2 = \left(\frac{1 + \sqrt{5}}{2}\right)^2 = \frac{1 + 2 \sqrt{5} + 5}{4} = \frac{6 + 2 \sqrt{5}}{4} = \frac{3 + \sqrt{5}}{2} = \frac{1 + \sqrt{5}}{2} + 1 = \phi + 1\).
\(\hat{\phi}^2 = \left(\frac{1  \sqrt{5}}{2}\right)^2 = \frac{1  2 \sqrt{5} + 5}{4} = \frac{6  2 \sqrt{5}}{4} = \frac{3  \sqrt{5}}{2} = \frac{1  \sqrt{5}}{2} + 1 = \hat{\phi} + 1\).
3.27
Prove by induction that the ith Fibonacci number satisfies the equality
\(F_i = \dfrac{\phi^i  \hat{\phi}^i}{\sqrt{5}}\),
where ϕ is the golden ratio and \(\hat{\phi}\) is its conjugate.
Base cases:
 If i = 0, \(\frac{\phi^i  \hat{\phi}^i}{\sqrt{5}} = \frac{1  1}{\sqrt{5}} = 0\), the claim holds.
 If i = 1, \(\frac{\phi^i  \hat{\phi}^i}{\sqrt{5}} = \frac{\phi  \hat{\phi}}{\sqrt{5}} = \frac{\frac{1 + \sqrt{5}}{2}  \frac{1  \sqrt{5}}{2}}{\sqrt{5}} = \frac{\sqrt{5}}{\sqrt{5}} = 1\), the claim holds.
Inductive case:
 By induction, we have \(F_{i  2} = \frac{\phi^{i  2}  \hat{\phi}^{i  2}}{\sqrt{5}}\) and \(F_{i  1} = \frac{\phi^{i  1}  \hat{\phi}^{i  1}}{\sqrt{5}}\). So \(F_i = F_{i  2} + F_{i  1} = \frac{\phi^{i  2}  \hat{\phi}^{i  2}}{\sqrt{5}} + \frac{\phi^{i  1}  \hat{\phi}^{i  1}}{\sqrt{5}} = \frac{\phi^{i  2} (1 + \phi)  \hat{\phi}^{i  2} \left(1 + \hat{\phi}\right)}{\sqrt{5}}\). Base on the conclusion of exercise 3.26, we have \(1 + \phi = \phi^2\) and \(1 + \hat{\phi} = \hat{\phi}^2\), so \(F_i = \frac{\phi^{i  2} \phi^2  \hat{\phi}^{i  2} \hat{\phi}^2}{\sqrt{5}} = \frac{\phi^i  \hat{\phi}^i}{\sqrt{5}}\), the claim holds.
3.28
Show that k ln k = Θ(n) implies k = Θ(n / ln n).
Skipped.
3.X Problems
31 Asymptotic behavior of polynomials
Let
p(n) = \(\displaystyle ∑_{i=0}^d a_i n^i\),
where \(a_d\) > 0, be a degreed polynomial in n, and let k be a constant. Use the definitions of the asymptotic notations to prove the following properties.
 If k ≥ d, then p(n) = O(\(n^k\)).
 If k ≤ d, then p(n) = Ω(\(n^k\)).
 If k = d, then p(n) = Θ(\(n^k\)).
 If k > d, then p(n) = o(\(n^k\)).
 If k < d, then p(n) = ω(\(n^k\)).
Skipped.
32 Relative asymptotic growths
Indicate, for each pair of expressions (A, B) in the table below, whether A is O, o, Ω, ω, or Θ of B. Assume that k ≥ 1, ϵ > 0, and c > 1 are constants. Your answer should be in the form of the table with “yes” or “no” written in each box.
A B O o Ω ω Θ a. \(\lg^k n\) \(n^ϵ\) b. \(n^k\) \(c^n\) c. \(\sqrt{n}\) \(n^{\sin n}\) d. \(2^n\) \(2^{n / 2}\) e. \(n^{\lg c}\) \(c^{\lg n}\) f. lg(n!) \(\lg\left(n^n\right)\)
A  B  O  o  Ω  ω  Θ  

a.  \(\lg^k n\)  \(n^ϵ\)  yes  yes  no  no  no 
b.  \(n^k\)  \(c^n\)  yes  yes  no  no  no 
c.  \(\sqrt{n}\)  \(n^{\sin n}\)  no  no  no  no  no 
d.  \(2^n\)  \(2^{n / 2}\)  no  no  yes  yes  no 
e.  \(n^{\lg c}\)  \(c^{\lg n}\)  yes  no  yes  no  yes 
f.  lg(n!)  \(\lg\left(n^n\right)\)  yes  no  yes  no  yes 
33 Ordering by asymptotic growth rates
Rank the following functions by order of growth; that is, find an arrangement \(g_1\), \(g_2\), …, \(g_{30}\) of the functions satisfying \(g_1\) = Ω(\(g_2\)), \(g_2\) = Ω(\(g_3\)), …, \(g_{29}\) = Ω(\(g_{30}\)). Partition your list into equivalence classes such that functions f(n) and g(n) are in the same class if and only if f(n) = Θ(g(n)).
\(\lg\left(\lg^* n\right)\) \(2^{\lg^* n}\) \(\left(\sqrt{2}\right)^{\lg n}\) \(n^2\) n! lg n! \(\left(\frac{3}{2}\right)^n\) \(n^3\) \(\lg^2 n\) lg(n!) \(2^{2^n}\) \(n^{1 / \lg n}\) ln ln n \(\lg^* n\) n ⋅ \(2^n\) \(n^{\lg \lg n}\) ln n 1 \(2^{\lg n}\) \(\left(\lg n\right)^{\lg n}\) \(e^n\) \(4^{\lg n}\) (n + 1)! \(\sqrt{\lg n}\) \(\lg^*\left(\lg n\right)\) \(2^{\sqrt{2 \lg n}}\) n \(2^n\) n lg n \(2^{2^{n + 1}}\) Give an example of a single nonnegative function f(n) such that for all functions \(g_i(n)\) in part (a), f(n) is neither \(O\left(g_i(n)\right)\) nor \(Ω\left(g_i(n)\right)\).
Skipped.
34 Asymptotic notation properties
Let f(n) and g(n) be asymptotically positive functions. Prove or disprove each of the following conjectures.
 f(n) = O(g(n)) implies g(n) = O(f(n)).
 f(n) + g(n) = Θ(min(f(n), g(n))).
 f(n) = O(g(n)) implies lg(f(n)) = O(lg(g(n))), where lg(g(n)) ≥ 1 and f(n) ≥ 1 for all sufficiently large n.
 f(n) = O(g(n)) implies \(2^{f(n)} = O\left(2^{g(n)}\right)\).
 f(n) = O(\((f(n))^2\)).
 f(n) = O(g(n)) implies g(n) = Ω(f(n)).
 f(n) = Θ(f(n / 2)).
 f(n) + o(f(n)) = Θ(f(n)).
Too lazy to prove. Just list my guessing here.
 False.
 False.
 Not sure.
 Not sure.
 Not sure.
 True.
 False.
 True.
35 Variations on O and Ω
Some authors define Ω in a slightly different way than we do; let’s use \(\overset{∞}{Ω}\) (read “omega infinity”) for this alternative definition. We say that f(n) = \(\overset{∞}{Ω}\)(g(n)) if there exists a positive constant c such that f(n) ≥ c g(n) ≥ 0 for infinitely many integers n.
 Show that for any two functions f(n) and g(n) that are asymptotically nonnegative, either f(n) = O(g(n)) or f(n) = \(\overset{∞}{Ω}\)(g(n)) or both, whereas this is not true if we use Ω in place of \(\overset{∞}{Ω}\).
 Describe the potential advantages and disadvantages of using \(\overset{∞}{Ω}\) instead of Ω to characterize the running times of programs.
Some authors also define O in a slightly different manner; let’s use O′ for the alternative definition. We say that f(n) = O′(g(n)) if and only if f(n) = O(g(n)).
 What happens to each direction of the “if and only if” in Theorem 3.1 if we substitute O′ for O but still use Ω?
Some authors define \(\widetilde{O}\) (read “softoh”) to mean O with logarithmic factors ignored:
\(\widetilde{O}(g(n))\) = {f(n) : there exist positive constants c, k, and \(n_0\) such that \(0 ≤ f (n) ≤ c g(n) \lg^k(n)\) for all \(n ≥ n_0\)}.
 Define \(\widetilde{Ω}\) and \(\widetilde{Θ}\) in a similar manner. Prove the corresponding analog to Theorem 3.1.
Skipped.
36 Iterated functions
We can apply the iteration operator \(^\ast\) used in the \(\lg^\ast\) function to any monotonically increasing function f(n) over the reals. For a given constant c ∈ ℝ, we define the iterated function \(f_c^*\) by
\(f_c^*(n) = \min \lbrace i ≥ 0 : f^{(i)}(n) ≤ c \rbrace\),
which need not be well defined in all cases. In other words, the quantity \(f_c^*(n)\) is the number of iterated applications of the function f required to reduce its argument down to c or less.
For each of the following functions f(n) and constants c, give as tight a bound as possible on \(f_c^*(n)\).
f(n)  c  \(f_c^*(n)\)  

a.  n  1  0  ⌈n⌉ 
b.  lg n  1  Don’t know 
c.  n / 2  1  ⌈lg n⌉ 
d.  n / 2  2  ⌈lg n  1⌉ 
e.  \(\sqrt{n}\)  2  ⌈lg lg n⌉ 
f.  \(\sqrt{n}\)  1  ∞ 
g.  \(n^{1 / 3}\)  2  ⌈\(\log_3 \lg n\)⌉ 
h.  n / lg n  2  Don’t know 
4 DivideandConquer
4.1 The maximumsubarray problem
4.11
What does FindMaximumSubarray return when all elements of A are negative?
An array containing the single maximum element of the original array.
4.12
Write pseudocode for the bruteforce method of solving the maximumsubarray problem. Your procedure should run in Θ(\(n^2\)) time.
FindMaximumSubarrayBruteForce(A)
 maxleft = 0
 maxright = 0
 maxsum = ∞
 for i = 1 to A.Length
 sum = 0
 for j = i to A.Length
 sum = sum + A[j]
 if sum > maxsum
 maxleft = i
 maxright = j
 maxsum = sum
 return (maxleft, maxright, maxsum)
4.13
Implement both the bruteforce and recursive algorithms for the maximumsubarray problem on your own computer. What problem size \(n_0\) gives the crossover point at which the recursive algorithm beats the bruteforce algorithm? Then, change the base case of the recursive algorithm to use the bruteforce algorithm whenever the problem size is less than \(n_0\). Does that change the crossover point?
See here for recursive implementation.
See here for bruteforce implementation.
Skipped crossover point test for now.
4.14
Suppose we change the definition of the maximumsubarray problem to allow the result to be an empty subarray, where the sum of the values of an empty subarray is 0. How would you change any of the algorithms that do not allow empty subarrays to permit an empty subarray to be the result?
Run the original algorithm first, if the maximum sum is negative, return an empty subarray.
4.15
Use the following ideas to develop a nonrecursive, lineartime algorithm for the maximumsubarray problem. Start at the left end of the array, and progress toward the right, keeping track of the maximum subarray seen so far. Knowing a maximum subarray of A[1‥j], extend the answer to find a maximum subarray ending at index j + 1 by using the following observation: a maximum subarray of A[1‥j + 1] is either a maximum subarray of A[1‥j] or a subarray A[i‥j + 1], for some 1 ≤ i ≤ j + 1. Determine a maximum subarray of the form A[i‥j + 1] in constant time based on knowing a maximum subarray ending at index j.
See here for implementation.
4.2 Strassen’s algorithm for matrix multiplication
4.21
Use Strassen’s algorithm to compute the matrix product
\(\begin{pmatrix}1 & 3\\7 & 5\end{pmatrix}\begin{pmatrix}6 & 8\\4 & 2\end{pmatrix}\).
Show your work.
Skipped.
4.22
Write pseudocode for Strassen’s algorithm.
Skipped.
4.23
How would you modify Strassen’s algorithm to multiply n × n matrices in which n is not an exact power of 2? Show that the resulting algorithm runs in time Θ(\(n^{\lg 7}\)).
Skipped.
4.24
What is the largest k such that if you can multiply 3 × 3 matrices using k multiplications (not assuming commutativity of multiplication), then you can multiply n × n matrices in time o(\(n^{\lg 7}\))? What would the running time of this algorithm be?
Skipped.
4.25
V. Pan has discovered a way of multiplying 68 × 68 matrices using 132,464 multiplications, a way of multiplying 70 × 70 matrices using 143,640 multiplications, and a way of multiplying 72 × 72 matrices using 155,424 multiplications. Which method yields the best asymptotic running time when used in a divideandconquer matrixmultiplication algorithm? How does it compare to Strassen’s algorithm?
Skipped.
4.26
How quickly can you multiply a k n × n matrix by an n × k n matrix, using Strassen’s algorithm as a subroutine? Answer the same question with the order of the input matrices reversed.
Skipped.
4.27
Show how to multiply the complex numbers a + b i and c + d i using only three multiplications of real numbers. The algorithm should take a, b, c, and d as input and produce the real component a c  b d and the imaginary component a d + b c separately.
Reference: (a + b i) × (c + d i) = a c  b d + (a d + b c) i.
Let w = (a + b) × (c  d), x = a × d, y = b × c, which uses three real number multiplications, then the result can be calculated as (w + x  y) + (x + y) i.
4.3 The substitution method for solving recurrences
4.31
Show that the solution of T(n) = T(n  1) + n is O(\(n^2\)).
T(n)
= T(n  1) + n
≤ c \((n  1)^2\) + n
= c \(n^2\)  2 c n + c + n
= c \(n^2\)  (2 c  1) n + c
Here we choose c = 1, we have
c \(n^2\)  (2 c  1) n + c = \(n^2\)  n + 1.
If n ≥ 1, we have \(n^2\)  n + 1 ≤ \(n^2\), that is, T(n) ≤ c \(n^2\), since c = 1.
4.32
Show that the solution of T(n) = T(⌈n / 2⌉) + 1 is O(lg n).
T(n)
= T(⌈n / 2⌉) + 1
≤ c lg ⌈n / 2⌉ + 1
= c lg (⌈n / 2⌉ \(2^{1 / c}\))
< c lg ((n / 2 + 1) \(2^{1 / c}\))
= c lg (\(2^{1 / c  1}\) n + \(2^{1 / c}\))
If we choose c = 2, we have
c lg (\(2^{1 / c  1}\) n + \(2^{1 / c}\)) = 2 lg (\(2^{1 / 2}\) n + \(2^{1 / 2}\)).
If n ≥ 2 \(\sqrt{2}\) + 2, we have \(2^{1 / 2}\) n + \(2^{1 / 2}\) ≤ n.
So for c = 2 and n ≥ 2 \(\sqrt{2}\) + 2, we have T(n) < c lg (\(2^{1 / c  1}\) n + \(2^{1 / c}\)) ≤ c lg n.
4.33
We saw that the solution of T(n) = 2 T(⌊n / 2⌋) + n is O(n lg n). Show that the solution of this recurrence is also Ω(n lg n). Conclude that the solution is Θ(n lg n).
T(n)
= 2 T(⌊n / 2⌋) + n
≥ 2 c ⌊n / 2⌋ lg ⌊n / 2⌋ + n
> 2 c (n / 2  1) lg (n / 2  1) + n
Skipped.
4.34
Show that by making a different inductive hypothesis, we can overcome the difficulty with the boundary condition T(1) = 1 for recurrence (4.19) without adjusting the boundary conditions for the inductive proof.
Skipped.
4.35
Show that Θ(n lg n) is the solution to the “exact” recurrence (4.3) for merge sort.
Skipped.
4.36
Show that the solution to T(n) = 2 T(⌊n / 2⌋ + 17) + n is O(n lg n).
T(n)
= 2 T(⌊n / 2⌋ + 17) + n
≤ 2 c (⌊n / 2⌋ + 17) lg (⌊n / 2⌋ + 17) + n
< 2 c ((n + 2) / 2 + 17) lg ((n + 2) / 2 + 17) + n
Skipped.
4.37
Using the master method in Section 4.5, you can show that the solution to the recurrence T(n) = 4 T(n / 3) + n is T(n) = Θ(\(n^{\log_3 4}\)). Show that a substitution proof with the assumption T(n) ≤ c \(n^{\log_3 4}\) fails. Then show how to subtract off a lowerorder term to make a substitution proof work.
Suppose T(n) = c \(n^{\log_3 4}\)  3 n, we have:
T(n)
= 4 T(n / 3) + n
= 4 (c \((n / 3)^{\log_3 4}\)  3 (n / 3)) + n
= 4 (c \(n^{\log_3 4}\) / 4  n) + n
= c \(n^{\log_3 4}\)  3 n.
That is exactly what we want.
4.38
Using the master method in Section 4.5, you can show that the solution to the recurrence T(n) = 4 T(n / 2) + n is T(n) = Θ(\(n^2\)). Show that a substitution proof with the assumption T(n) ≤ c \(n^2\) fails. Then show how to subtract off a lowerorder term to make a substitution proof work.
Suppose T(n) = c \(n^2\)  n, we have
T(n)
= 4 (c \((n / 2)^2\)  n / 2) + n
= 4 (c \(n^2\) / 4  n / 2) + n
= c \(n^2\)  2 n + n
= c \(n^2\)  n.
That is exactly what we want.
4.39
Solve the recurrence T(n) = 3 T(\(\sqrt{n}\)) + log n by making a change of variables. Your solution should be asymptotically tight. Do not worry about whether values are integral.
Let m = log n, we have n = \(10^m\), so
T(\(10^m\))
= 3 T(\(\sqrt{10^m}\)) + log \(10^m\)
= 3 T(\(10^{m / 2}\)) + m.
Let S(m) = T(\(10^m\)), we have:
S(m) = 3 S(m / 2) + m.
I guess S(m) = c \(m^{\lg 3}\)  2 m,
S(m)
= 3 (c \((m / 2)^{\lg 3}\)  2 (m / 2)) + m
= 3 (c \(m^{\lg 3}\) / 3  m) + m
= c \(m^{\lg 3}\)  3 m + m
= c \(m^{\lg 3}\)  2 m.
So my guess is right.
T(n)
= S(log n)
= c \((\log n)^{\lg 3}\)  2 log n.
Notes:
Solving T(n) = a T(n / b) + k \(n^p\): if p = \(\log_b a\), T(n) = k \(n^p \log_b n\) + c \(n^p\), otherwise T(n) = c \(n^{\log_b a}\) + (k / (1  a / \(b^p\))) \(n^p\).
4.4 The recursiontree method for solving recurrences
4.41
Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = 3 T(⌊n / 2⌋) + n. Use the substitution method to verify your answer.
T(n) ≤ c \(n^{\lg 3}\)  2 n.
Verification:
T(n)
= 3 T(⌊n / 2⌋) + n
≤ 3 (c \(\left\lfloor{}n / 2\right\rfloor^{\lg 3}\)  2 ⌊n / 2⌋) + n
≤ 3 (c \((n / 2)^{\lg 3}\)  2 (n / 2)) + n
= 3 (c \(n^{\lg 3}\) / 3  n) + n
= 3 c \(n^{\lg 3}\) / 3  3 n + n
= c \(n^{\lg 3}\)  2 n.
4.42
Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = T(n / 2) + \(n^2\). Use the substitution method to verify your answer.
T(n) = c + (4 / 3) \(n^2\).
Verification:
T(n)
= c + (4 / 3) \((n / 2)^2\) + \(n^2\)
= c + (1 / 3) \(n^2\) + \(n^2\)
= c + (4 / 3) \(n^2\).
4.43
Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = 4 T(n / 2 + 2) + n. Use the substitution method to verify your answer.
T(n) = c \(n^2\)  (8 c + 1) n + 16 c + 8 / 3.
Verification:
T(n)
= 4 T(n / 2 + 2) + n
= 4 (c \((n / 2 + 2)^2\)  (8 c + 1) (n / 2 + 2) + 16 c + 8 / 3) + n
= 4 (c (\(n^2\) / 4 + 2 n + 4)  (4 c n + 16 c + n / 2 + 2) + 16 c + 8 / 3) + n
= 4 (c \(n^2\) / 4  (2 c + 1 / 2) n + 4 c + 2 / 3) + n
= c \(n^2\)  (8 c + 2) n + 16 c + 8 / 3 + n
= c \(n^2\)  (8 c + 1) n + 16 c + 8 / 3.
4.44
Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = 2 T(n  1) + 1. Use the substitution method to verify your answer.
T(n) = c \(2^n\)  1
Verification:
T(n)
= 2 T(n  1) + 1
= 2 (c \(2^{n  1}\)  1) + 1
= c \(2^n\)  2 + 1
= c \(2^n\)  1.
4.45
Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = T(n  1) + T(n / 2) + n. Use the substitution method to verify your answer.
T(n) = O(\(2^n\)), and T(n) = Ω(n lg n).
Skipped.
4.46
Argue that the solution to the recurrence T(n) = T(n / 3) + T(2 n / 3) + c n, where c is a constant, is Ω(n lg n) by appealing to a recursion tree.
On each level of recursion whose depth is less than lg n / lg 3, the cost on this level is c n, so the cost is at least c n lg n / lg 3, that is T(n) = Ω(n lg n).
4.47
Draw the recursion tree for T(n) = 4 T(⌊n / 2⌋) + c n, where c is a constant, and provide a tight asymptotic bound on its solution. Verify your bound by the substitution method.
T(n) ≥ k \(n^2\)  (c  4 k) n + 4 k  4 c / 3.
Verification:
T(n)
= 4 T(⌊n / 2⌋) + c n
≥ 4 T(n / 2  1) + c n
≥ 4 (k \((n / 2  1)^2\)  (c  4 k) (n / 2  1) + 4 k  4 c / 3) + c n
= k \(n^2\)  (c  4 k) n + 4 k  4 c / 3.
T(n) ≤ k \(n^2\)  c n.
Verification:
T(n)
= 4 T(⌊n / 2⌋) + c n
≤ 4 T(n / 2) + c n
≤ 4 (k \((n / 2)^2\)  c (n / 2)) + c n
= k \(n^2\)  c n.
So T(n) = Θ(\(n^2\)).
4.48
Use a recursion tree to give an asymptotically tight solution to the recurrence T(n) = T(n  a) + T(a) + c n, where a ≥ 1 and c > 0 are constants.
T(n) = (c / (2 a)) \(n^2\) + k n  a c.
Verification:
T(n)
= T(n  a) + T(a) + c n
= (c / (2 a)) \((n  a)^2\) + k (n  a)  a c + (c / (2 a)) \(a^2\) + k a  a c + c n
= (c / (2 a)) (\(n^2\)  2 a n + \(a^2\)) + k n  a k  2 a c + a c / 2 + k a + c n
= (c / (2 a)) \(n^2\)  c n + a c / 2 + k n  2 a c + a c / 2 + c n
= (c / (2 a)) \(n^2\) + k n  a c
4.49
Use a recursion tree to give an asymptotically tight solution to the recurrence T(n) = T(α n) + T((1  α) n) + c n, where α is a constant in the range 0 < α < 1 and c > 0 is also a constant.
Like exercise 4.46, we can prove T(n) = Ω(n lg n), and T(n) = O(n lg n), so T(n) = Θ(n lg n).
4.5 The master method for solving recurrences
4.51
Use the master method to give tight asymptotic bounds for the following recurrences.
 T(n) = 2 T(n / 4) + 1.
 T(n) = 2 T(n / 4) + \(\sqrt{n}\).
 T(n) = 2 T(n / 4) + n.
 T(n) = 2 T(n / 4) + \(n^2\).
 T(n) = Θ(\(\sqrt{n}\)).
 T(n) = Θ(\(\sqrt{n}\) lg n).
 T(n) = Θ(n).
 T(n) = Θ(\(n^2\)).
4.52
Professor Caesar wishes to develop a matrixmultiplication algorithm that is asymptotically faster than Strassen’s algorithm. His algorithm will use the divideandconquer method, dividing each matrix into pieces of size n / 4 × n / 4, and the divide and combine steps together will take Θ(\(n^2\)) time. He needs to determine how many subproblems his algorithm has to create in order to beat Strassen’s algorithm. If his algorithm creates a subproblems, then the recurrence for the running time T(n) becomes T(n) = a T(n / 4) + Θ(\(n^2\)). What is the largest integer value of a for which Professor Caesar’s algorithm would be asymptotically faster than Strassen’s algorithm?
The answer is 48. Since \(log_4 49\) = lg 7.
4.53
Use the master method to show that the solution to the binarysearch recurrence T(n) = T(n / 2) + Θ(1) is T(n) = Θ(lg n). (See Exercise 2.35 for a description of binary search.)
Since Θ(1) = Θ(\(n^{\log_2 1}\)), T(n) = Θ(\(n^{\log_2 1}\) lg n) = Θ(lg n).
4.54
Can the master method be applied to the recurrence T(n) = 4 T(n / 2) + \(n^2\) lg n? Why or why not? Give an asymptotic upper bound for this recurrence.
The master method can not be applied to that recurrence, because \(n^2\) lg n is asymptotically large than \(n^{\log_2 4} = n^2\), but it is not polynomially asymptotically large than \(n^{\log_2 4}\).
T(n) = \(n^2 \left(\lg^2 n + \lg n + c\right)\) / 2.
Verification:
T(n)
= 4 T(n / 2) + \(n^2\) lg n
= 4 \((n / 2)^2 \left(\lg^2 (n / 2) + \lg (n / 2) + c\right)\) / 2 + \(n^2\) lg n
= 4 \((n / 2)^2 \left((\lg n  1)^2 + \lg n  1 + c\right)\) / 2 + \(n^2\) lg n
= 4 \((n / 2)^2 \left(\lg^2 n  2 \lg n + 1 + \lg n  1 + c\right)\) / 2 + \(n^2\) lg n
= \(n^2 \left(\lg^2 n  \lg n + c\right)\) / 2 + \(n^2\) lg n
= \(n^2 \left(\lg^2 n + \lg n + c\right)\) / 2.
4.55 ★
Consider the regularity condition a f(n / b) ≤ c f(n) for some constant c < 1, which is part of case 3 of the master theorem. Give an example of constants a ≥ 1 and b > 1 and a function f(n) that satisfies all the conditions in case 3 of the master theorem except the regularity condition.
Skipped.
4.6 Proof of the master theorem ★
4.61 ★
Give a simple and exact expression for \(n_j\) in equation (4.27) for the case in which b is a positive integer instead of an arbitrary real number.
Theorem 3.4 and 3.5:
For any real number x ≥ 0 and integers a, b > 0:
 \(\left\lceil\frac{\left\lceil x / a\right\rceil}{b}\right\rceil = \left\lceil\frac{x}{a b}\right\rceil\),
 \(\left\lfloor\frac{\left\lfloor x / a\right\rfloor}{b}\right\rfloor = \left\lfloor\frac{x}{a b}\right\rfloor\).
\(n_j\) = ⌈n / \(b^j\)⌉.
Proof by induction:
 If j = 0, \(n_j\) = ⌈n / \(b^j\)⌉ = ⌈n / \(b^0\)⌉ = ⌈n⌉ = n, the claim holds.
 If j > 0, \(n_j\) = ⌈\(n_{j  1}\) / b⌉ = ⌈⌈n / \(b^{j  1}\)⌉ / b⌉, since both b and \(b^{j  1}\) are integers, ⌈⌈n / \(b^{j  1}\)⌉ / b⌉ = ⌈n / \(b^{j  1}\) / b⌉ = ⌈n / \(b^j\)⌉. So \(n_j\) = ⌈n / \(b^j\)⌉, the claim holds.
Notes:
In the book Concrete Mathematics, there is a theorem:

Let f(x) be any continuous, monotonically increasing function with the property that
 f(x) = integer ⇒ x = integer.
Then we have
 ⌊f(x)⌋ = ⌊f(⌊x⌋)⌋ and ⌈f(x)⌉ = ⌈f(⌈x⌉)⌉,
whenever f(x), f(⌊x⌋), and f(⌈x⌉) are defined.
Proof:
f(⌊x⌋)  1 < ⌊f(⌊x⌋)⌋ ≤ f(⌊x⌋)
⇒ ⌊f(⌊x⌋)⌋ ≤ f(⌊x⌋) < ⌊f(⌊x⌋)⌋ + 1
⇒ \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor\right)\)
≤ \(f^{1}\left(f\left(\left\lfloor x\right\rfloor\right)\right)\)
< \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\)
⇒ \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor\right)\)
≤ ⌊x⌋
< \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\).
Since ⌊f(⌊x⌋)⌋ + 1 is an integer, and we also have f(x) = integer ⇒ x = integer, we know that \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\) is an integer. Because ⌊x⌋ < \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\), we have x < \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\). So:
\(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor\right)\)
≤ ⌊x⌋
< \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\)
⇒ \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor\right)\)
≤ x
< \(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\)
⇒ \(f\left(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor\right)\right)\)
≤ f(x)
< \(f\left(f^{1}\left(\left\lfloor f\left(\left\lfloor x\right\rfloor\right)\right\rfloor + 1\right)\right)\)
⇒ ⌊f(⌊x⌋)⌋ ≤ f(x) < ⌊f(⌊x⌋)⌋ + 1
⇒ f(x) = ⌊f(⌊x⌋)⌋.
The same method can be applied to proving ⌈f(x)⌉ = ⌈f(⌈x⌉)⌉.
4.62 ★
Show that if f(n) = Θ(\(n^{\log_b a} \lg^k n\)), where k ≥ 0, then the master recurrence has solution T(n) = Θ(\(n^{\log_b a} \lg^{k + 1} n\)). For simplicity, confine your analysis to exact powers of b.
T(n)
= Θ(\(n^{\log_b a}\)) + \(∑_{j = 0}^{\log_b n  1} a^j f(n / b^j)\)
= Θ(\(n^{\log_b a}\)) + \(∑_{j = 0}^{\log_b n  1} a^j Θ\left(\left(n / b^j\right)^{\log_b a} \lg^k \left(n / b^j\right)\right)\)
= Θ(\(n^{\log_b a}\)) + \(∑_{j = 0}^{\log_b n  1} Θ\left(a^j \left(n / b^j\right)^{\log_b a} \lg^k \left(n / b^j\right)\right)\)
= Θ(\(n^{\log_b a}\)) + \(∑_{j = 0}^{\log_b n  1} Θ\left(n^{\log_b a} \left(\lg n  \lg \left(b^j\right)\right)^k\right)\)
= Θ(\(n^{\log_b a}\)) + \(∑_{j = 0}^{\log_b n  1} Θ\left(n^{\log_b a} \left(\lg n  j \lg b\right)^k\right)\)
= Θ(\(n^{\log_b a}\)) + \(∑_{j = 0}^{\log_b n  1} Θ\left(n^{\log_b a} \lg^k n\right)\)
= Θ(\(n^{\log_b a}\)) + \(n^{\log_b a} Θ\left(\left(\log_b n  1\right)\lg^k n\right)\)
= Θ(\(n^{\log_b a}\)) + \(n^{\log_b a} Θ\left(\log_b n \lg^k n  \lg^k n\right)\)
= Θ(\(n^{\log_b a}\)) + \(n^{\log_b a} Θ\left(\lg^{k + 1} n\right)\)
= Θ(\(n^{\log_b a} \lg^{k + 1} n\)).
4.63 ★
Show that case 3 of the master theorem is overstated, in the sense that the regularity condition a f(n / b) ≤ c f(n) for some constant c < 1 implies that there exists a constant ϵ > 0 such that f(n) = Ω(\(n^{\log_b a + ϵ}\)).
Let ϵ = \(\log_b c\), I guess f(n) = Ω(\(n^{\log_b a  \log_b c}\)).
f(n)
≥ (a / c) f(n / b)
= (a / c) Ω(\((n / b)^{\log_b a  \log_b c}\))
= (a / c) Ω(\(n^{\log_b a  \log_b c} / (a / c)\))
= Ω(\(n^{\log_b a  \log_b c}\)).
4.X Problems
41 Recurrence examples
Give asymptotic upper and lower bounds for T(n) in each of the following recurrences. Assume that T(n) is constant for n ≤ 2. Make your bounds as tight as possible, and justify your answers.
 T(n) = 2 T(n / 2) + \(n^4\).
 T(n) = T(7 n / 10) + n.
 T(n) = 16 T(n / 4) + \(n^2\).
 T(n) = 7 T(n / 3) + \(n^2\).
 T(n) = 7 T(n / 2) + \(n^2\).
 T(n) = 2 T(n / 4) + \(\sqrt{n}\)
 T(n) = T(n  2) + \(n^2\).
 T(n) = c n + (8 / 7) \(n^4\).
 T(n) = (10 / 3) n + c.
 T(n) = \(n^2 \log_4 n\) + c \(n^2\).
 T(n) = (9 / 2) \(n^2\) + c \(n^{\log_3 7}\).
 T(n) = c \(n^{\lg 7}\)  (4 / 3) \(n^2\).
 T(n) = \(\sqrt{n} \log_4 n\) + c \(\sqrt{n}\).
 T(n) = \(n^3\) / 6 + \(n^2\) / 2 + n / 3 + c.
42 Parameterpassing costs
Throughout this book, we assume that parameter passing during procedure calls takes constant time, even if an Nelement array is being passed. This assumption is valid in most systems because a pointer to the array is passed, not the array itself. This problem examines the implications of three parameterpassing strategies:
 An array is passed by pointer. Time = Θ(1).
 An array is passed by copying. Time = Θ(N), where N is the size of the array.
 An array is passed by copying only the subrange that might be accessed by the called procedure. Time = Θ(q  p + 1) if the subarray A[p‥q] is passed.
 Consider the recursive binary search algorithm for finding a number in a sorted array (see Exercise 2.35). Give recurrences for the worstcase running times of binary search when arrays are passed using each of the three methods above, and give good upper bounds on the solutions of the recurrences. Let N be the size of the original problem and n be the size of a subproblem.
 Redo part (a) for the MergeSort algorithm from Section 2.3.1.
 Binary search algorithm

By pointer: T(n) = T(n / 2) + Θ(1).
T(N) = Θ(lg N).

By copying: T(n) = T(n / 2) + Θ(N).
T(N) = Θ(N lg N).

By copying subrange: T(n) = T(n / 2) + Θ(n).
T(N) = Θ(N).

 Merge sort algorithm

By pointer: T(n) = 2 T(n / 2) + Θ(n).
T(N) = Θ(N lg N).

By copying: T(n) = 2 T(n / 2) + Θ(n) + Θ(N*).
T(n)
= Θ(n) + \(∑_{j = 0}^{\lg n  1} Θ\left(2^j (n / 2^j)\right)\) + \(∑_{j = 0}^{\lg n  1} Θ\left(2^j N\right)\)
= Θ(n) + Θ(n lg n) + Θ(N n)T(N) = Θ(\(N^2\)).

By copying subrange: T(n) = 2 T(n / 2) + Θ(n), T(N) = Θ(N lg N).

43 More recurrence examples
Give asymptotic upper and lower bounds for T(n) in each of the following recurrences. Assume that T(n) is constant for sufficiently small n. Make your bounds as tight as possible, and justify your answers.
 T(n) = 4 T(n / 3) + n lg n.
 T(n) = 3 T(n / 3) + n / lg n.
 T(n) = 4 T(n / 2) + \(n^2 \sqrt{n}\).
 T(n) = 3 T(n / 3  2) + n / 2.
 T(n) = 2 T(n / 2) + n / lg n.
 T(n) = T(n / 2) + T(n / 4) + T(n / 8) + n.
 T(n) = T(n  1) + 1 / n.
 T(n) = T(n  1) + lg n.
 T(n) = T(n  2) + 1 / lg n.
 T(n) = \(\sqrt{n}\) T(\(\sqrt{n}\)) + n
 T(n) = c \(n^{\log_3 4}\)  3 n lg n  12 n lg 3.
 T(n) = Θ(n), according to exercise 4.62.
 T(n) = (2 + \(\sqrt{2}\)) \(n^{5 / 2}\) + c \(n^2\).
 T(n) = (1 / 2) n \(\log_3 (n + 3)\) + c n + (3 / 2) \(\log_3 (n + 3)\) + 3 c + 3 / 4.
 T(n) = Θ(n), according to exercise 4.62.
 T(n) = 8 n.
 T(n) = c + \(∑_{j = 2}^n (1 / j)\) = Θ(lg n).
 T(n) = c + \(∑_{j = 2}^n \lg n\) = c + \(\lg ∏_{j = 2}^n n\) = c + lg n! = Θ(n lg n).
 Skipped.
 T(n) = n lg lg n + c n.
44 Fibonacci numbers
This problem develops properties of the Fibonacci numbers, which are defined by recurrence (3.22). We shall use the technique of generating functions to solve the Fibonacci recurrence. Define the generating function (or formal power series) ℱ as
\(\begin{aligned} ℱ(z) &= \displaystyle ∑_{i = 0}^∞ F_i z^i\\ &= 0 + z + z^2 + 2 z^3 + 3 z^4 + 5 z^5 + 8 z^6 + 13 z^7 + 21 z^8 + ⋯, \end{aligned}\)
where \(F_i\) is the ith Fibonacci number.
Show that ℱ(z) = z + z ℱ(z) + \(z^2\) ℱ(z).
Show that
\(\begin{aligned} ℱ(z) &= \frac{z}{1  z  z^2}\\ &= \frac{z}{(1  \phi z)\left(1  \hat{\phi} z\right)}\\ &= \frac{1}{\sqrt{5}}\left(\frac{1}{1  \phi z}  \frac{1}{1  \hat{\phi} z}\right), \end{aligned}\)
where
ϕ = \(\dfrac{1 + \sqrt{5}}{2}\) = 1.61803…
and
\(\hat{\phi}\) = \(\dfrac{1  \sqrt{5}}{2}\) = 1.61803….
Show that
ℱ(z) = \(\displaystyle ∑_{i = 0}^∞ \frac{1}{\sqrt{5}}\left(\phi^i  \hat{\phi}^i\right) z^i\).
Use part (c) to prove that \(F_i = \phi^i / \sqrt{5}\) for i > 0, rounded to the nearest integer. (Hint: Observe that \(\hat{\phi}\) < 1.)

By definition:
\(\begin{aligned} z + z ℱ(z) + z^2 ℱ(z) &= z + z ∑_{i = 0}^∞ F_i z^i + z^2 ∑_{i = 0}^∞ F_i z^i\\ &= z + ∑_{i = 0}^∞ F_i z^{i + 1} + ∑_{i = 0}^∞ F_i z^{i + 2}\\ &= z + ∑_{i = 1}^∞ F_{i  1} z^i + ∑_{i = 2}^∞ F_{i  2} z^i\\ &= z + \left(F_0 z^1 + ∑_{i = 2}^∞ F_{i  1} z^i\right) + ∑_{i = 2}^∞ F_{i  2} z^i\\ &= z + F_0 z^1 + \left(∑_{i = 2}^∞ F_{i  1} z^i + ∑_{i = 2}^∞ F_{i  2} z^i\right)\\ &= F_1 z^1 + F_0 z^0 + ∑_{i = 2}^∞ F_i z^i\\ &= F_0 z^0 + F_1 z^1 + ∑_{i = 2}^∞ F_i z^i\\ &= ∑_{i = 0}^∞ F_i z^i\\ &= ℱ(z). \end{aligned}\)

ℱ(z) = z + z ℱ(z) + \(z^2\) ℱ(z)
⇒ ℱ(z)  z ℱ(z)  \(z^2\) ℱ(z) = z
⇒ ℱ(z) (1  z  \(z^2\)) = z
⇒ ℱ(z) = z / (1  z  \(z^2\)).z / ((1  ϕ z) (1  \(\hat{\phi}\) z))
= z / (1  (ϕ + \(\hat{\phi}\)) z + ϕ \(\hat{\phi} z^2\))
= z / (1  z  \(z^2\))
= ℱ(z).\(\frac{1}{\sqrt{5}} \left(\frac{1}{1  \phi z}  \frac{1}{1  \hat{\phi} z}\right)\)
= \(\frac{1}{\sqrt{5}} \frac{\left(1  \hat{\phi} z\right)  (1  \phi z)}{(1  \phi z) \left(1  \hat{\phi} z\right)}\)
= \(\frac{1}{\sqrt{5}} \frac{\left(\phi  \hat{\phi}\right) z}{(1  \phi z) \left(1  \hat{\phi} z\right)}\)
= \(\frac{1}{\sqrt{5}} \frac{\sqrt{5} z}{(1  \phi z) \left(1  \hat{\phi} z\right)}\)
= \(\frac{z}{(1  \phi z) \left(1  \hat{\phi} z\right)}\)
= ℱ(z).An interesting discovery: if we let z = 1, we have ℱ(1) = 1 / (1  1  \(1^2\)) = 1. Also, according to the definition of ℱ, we have: \(ℱ(1) = ∑_{i = 0}^∞ F_i 1^i = ∑_{i = 0}^∞ F_i\).
So we have \(∑_{i = 0}^∞ F_i\) = 0 + 1 + 1 + 2 + 3 + 5 + 8 + 13 + 21 + … = 1, WTF?

\(∑_{i = 0}^∞ \frac{1}{\sqrt{5}}\left(\phi^i  \hat{\phi}^i\right) z^i\)
= \(\frac{1}{\sqrt{5}} ∑_{i = 0}^∞ \left(\phi^i  \hat{\phi}^i\right) z^i\)
= \(\frac{1}{\sqrt{5}} \left(∑_{i = 0}^∞ (\phi z)^i  ∑_{i = 0}^∞ \left(\hat{\phi} z\right)^i\right)\)
= \(\frac{1}{\sqrt{5}} \left(\frac{1}{1  \phi z}  \frac{1}{1  \hat{\phi} z}\right)\)
= ℱ(z). 
\(F_i = \frac{1}{\sqrt{5}}\left(\phi^i  \hat{\phi}^i\right) = \phi^i / \sqrt{5}  \hat{\phi}^i / \sqrt{5}\)
⇒ \(F_i  \phi^i / \sqrt{5} =  \hat{\phi}^i / \sqrt{5}\).Since \(\left \hat{\phi}^i / \sqrt{5}\right\) < 0.5, and \(F_i\) is an integer, \(F_i = \phi^i / \sqrt{5}\), rounded to the nearest integer.
45 Chip testing
Professor Diogenes has n supposedly identical integratedcircuit chips that in principle are capable of testing each other. The professor’s test jig accommodates two chips at a time. When the jig is loaded, each chip tests the other and reports whether it is good or bad. A good chip always reports accurately whether the other chip is good or bad, but the professor cannot trust the answer of a bad chip. Thus, the four possible outcomes of a test are as follows:
Chip A says Chip B says Conclusion B is good A is good both are good, or both are bad B is good A is bad at least one is bad B is bad A is good at least one is bad B is bad A is bad at least one is bad
 Show that if at least n / 2 chips are bad, the professor cannot necessarily determine which chips are good using any strategy based on this kind of pairwise test. Assume that the bad chips can conspire to fool the professor.
 Consider the problem of finding a single good chip from among n chips, assuming that more than n / 2 of the chips are good. Show that ⌊n / 2⌋ pairwise tests are sufficient to reduce the problem to one of nearly half the size.
 Show that the good chips can be identified with Θ(n) pairwise tests, assuming that more than n / 2 of the chips are good. Give and solve the recurrence that describes the number of tests.

If there are at least n / 2 chips are bad, there must exist same number of bad chips from good chips, and the bad chips can simulate whatever the behavior the good chips have, so it is not possible to distinguish bad chips from good chips.

Generalize the original problem to this:
Assume there are no less good chip than bad chips:
 If the number of good chips is greater than the number of bad chips, find a good chip.
 If the number of good chips is greater than or equal to the number of bad chips, find a good chip or say the number of good chips equal to the number of bad chips.
 Otherwise, the result is undefined.
Solution:
 Base case: If there is zero chip, we say that the number of good chips equal to the number of bad chips.
 Inductive cases:
 If the number of chips is even, we group them into pairs, then in each pair, we test each chip with the other chip. If the chips in one pair both say the other one is a good chip, we throw away any one chip in this pair; otherwise, we throw away both chips. Then we will be left with at most half of the original chips. And since chip pairs that do not say each other is good ether have one bad chip or have two bad chips, throwing them away does not change the fact that good chips are not less than bad chips. The remaining chip pairs are either both good chips or bad chips, after throwing away one chip in every those pairs, we have reduced the size of the problem to at most half of the original problem size.
 If the number of chips is odd, we know that the number of good chip must be greater than the number of bad chips, since there can not be the same number of good chips and bad chips. We randomly remove one chip from the chips, and we will be left with even number of chips, in which good chips are no less than bad chips. We process the remaining chips with the method used in the even number case. After the remaining chips being processed, either we get a good chip, or we are told that the number of good chips are the same as the number of bad chips, which means the chip we removed is a good chip. Either way, we can get a good chip.
The solution to the generalized problem applies to the original problem.

With the solution above, we can find one good chip in number T(n) ≤ T(n / 2) + Θ(n) pair tests. According to the master theorem, we have T(n) = O(n). After we found one good we can identify all good chips with that good chip in Θ(n) time, so the total number of pairwise tests equals to O(n) + Θ(n) = Θ(n).
The solution is implemented here.
46 Monge arrays
An m × n array A of real numbers is a Monge array if for all i, j, k, and l such that 1 ≤ i < k ≤ m and 1 ≤ j < l ≤ n, we have
A[i, j] + A[k, l] ≤ A[i, l] + A[k, j].
In other words, whenever we pick two rows and two columns of a Monge array and consider the four elements at the intersections of the rows and the columns, the sum of the upperleft and lowerright elements is less than or equal to the sum of the lowerleft and upperright elements. For example, the following array is Monge:
10 17 13 28 23 17 22 16 29 23 24 28 22 34 24 11 13 6 17 7 45 44 32 37 23 36 33 19 21 6 75 66 51 53 34
Prove that an array is Monge if and only if for all i = 1, 2, …, m  1 and j = 1, 2, …, n  1, we have A[i, j] + A[i + 1, j + 1] ≤ A[i, j + 1] + A[i + 1, j].
(Hint: For the “if” part, use induction separately on rows and columns.)
The following array is not Monge. Change one element in order to make it Monge. (Hint: Use part (a).)
37 23 22 32 21 6 7 10 53 34 30 31 32 13 9 6 43 21 15 8 Let f(i) be the index of the column containing the leftmost minimum element of row i. Prove that f(1) ≤ f(2) ≤ ⋯ ≤ f(m) for any m × n Monge array.
Here is a description of a divideandconquer algorithm that computes the leftmost minimum element in each row of an m × n Monge array A:
 Construct a submatrix A′ of A consisting of the evennumbered rows of A. Recursively determine the leftmost minimum for each row of A′. Then compute the leftmost minimum in the oddnumbered rows of A.
Explain how to compute the leftmost minimum in the oddnumbered rows of A (given that the leftmost minimum of the evennumbered rows is known) in O(m + n) time.
Write the recurrence describing the running time of the algorithm described in part (d). Show that its solution is O(m + n log m).

Proof:
 The “only if” part is trivial.
 The “if” part:

First, we prove that: for all i, j, and l such that 1 ≤ i < m and 1 ≤ j < l ≤ n, we have A[i, j] + A[i + 1, l] ≤ A[i, l] + A[i + 1, j].
Proof by induction:

Base case: Trivially, A[i, j] + A[i + 1, j + 1] ≤ A[i, j + 1] + A[i + 1, j].

Inductive case: by induction, we have A[i, j] + A[i + 1, l  1] ≤ A[i, l  1] + A[i + 1, j]. Since A[i, l  1] + A[i + 1, l] ≤ A[i, l] + A[i + 1, l  1], we have:
(A[i, j] + A[i + 1, l  1]) + (A[i, l  1] + A[i + 1, l]) ≤ (A[i, l  1] + A[i + 1, j]) + (A[i, l] + A[i + 1, l  1])
⇒ (A[i, j] +A[i + 1, l  1]) + (A[i, l  1]+ A[i + 1, l]) ≤ (A[i, l  1]+ A[i + 1, j]) + (A[i, l] +A[i + 1, l  1])
⇒ A[i, j] + A[i + 1, l] ≤ A[i + 1, j] + A[i, l]
⇒ A[i, j] + A[i + 1, l] ≤ A[i, l] + A[i + 1, j].


Second, we prove the original claim by induction.

Base case: according to the first step, we have: for all i, j, and l such that 1 ≤ i < m and 1 ≤ j < l ≤ n, we have A[i, j] + A[i + 1, l] ≤ A[i, l] + A[i + 1, j].

Inductive case: by induction, A[i, j] + A[k  1, l] ≤ A[i, l] + A[k  1, j]. Since A[k  1, j] + A[k, l] ≤ A[k  1, l] + A[k, j], we have:
(A[i, j] + A[k  1, l]) + (A[k  1, j] + A[k, l]) ≤ (A[i, l] + A[k  1, j]) + (A[k  1, l] + A[k, j])
⇒ (A[i, j] +A[k  1, l]) + (A[k  1, j]+ A[k, l]) ≤ (A[i, l] +A[k  1, j]) + (A[k  1, l]+ A[k, j])
⇒ A[i, j] + A[k, l] ≤ A[i, l] + A[k, j]



Result:
37 23 24 32 21 6 7 10 53 34 30 31 32 13 9 6 43 21 15 8 
Proof by contradiction:
Assume for some i, f(i) > f(i + 1), let f(i) = j, and f(i + 1) = l. Since f(i) is the leftmost minimum element of row i, and f(i + 1) is the leftmost minimum element of row i + 1, we have
A[i, l] > A[i, j], and A[i + 1, j] > A[i + 1, l]
⇒ A[i, l] + A[i + 1, j] > A[i, j] + A[i + 1, l].That violates the condition of Monge array, so our assumption is wrong. That proves the claim is right.

If f(i  1) = j, and f(i + 1) = l, j ≤ f(i) ≤ l.
The time for computing the leftmost minimal element of odd row i is Θ(f(i + 1)  f(i  1) + 1).

If the number of rows is even, the time for computing the leftmost minimal element of all odd rows is
Θ(f(2) + \(∑_{i = 1}^{m / 2  1} (f(2 i + 2)  f(2 i) + 1)\))
= Θ(f(2) + (f(m)  f(2) + m / 2  1))
= Θ(f(m) + m / 2  1)
= O(n + m / 2  1)
= O(n + m) 
If the number of rows is odd, the time for computing the leftmost minimal element of all odd rows is:
Θ(f(2) + \(∑_{i = 1}^{(m  3) / 2} (f(2 i + 2)  f(2 i) + 1)\) + (n  f(m  1) + 1))
= Θ(f(2) + (f(m  1)  f(2) + (m  3) / 2) + (n  f(m  1) + 1))
= Θ((m  3) / 2 + n + 1)
= Θ(m / 2 + n  1 / 2)
= O(m + n)
So the time cost for computing the leftmost minimum in the oddnumbered rows is O(m + n).


T(m, n) = T(m / 2, n) + O(m + n).
T(m, n)
= Θ(1) + O(\(∑_{i=0}^{\lg m  1} \left(m / \left(2^i\right) + n\right)\))
= Θ(1) + O(\(∑_{i=0}^{\lg m  1} \left(m / \left(2^i\right)\right) + ∑_{i=0}^{\lg m  1} n\))
= Θ(1) + O(m \(∑_{i=0}^{\lg m  1} (1 / 2)^i + ∑_{i=0}^{\lg m  1} n\))
= Θ(1) + O(m (1  \((1 / 2)^{\lg m}\)) / (1  1 / 2) + n lg m)
= Θ(1) + O(2 m (1  (1 / m)) + n lg m)
= Θ(1) + O(2 m  2 + n lg m)
= O(m + n lg m)
5 Probabilistic Analysis and Randomized Algorithms
5.1 The hiring problem
5.11
Show that the assumption that we are always able to determine which candidate is best, in line 4 of procedure HireAssistant, implies that we know a total order on the ranks of the candidates.
We can prove it by proving its contrapositive:
 If the ranks of the candidates don’t form a total order, there exist a set of candidates in which we are not able to determine which candidate is best.
Proof:
 Total relation is violated: If there exist two candidates that we can not compare them, then we can not decide which one is best.
 Reflexive is violated: Skipped.
 Antisymmetric is violated: If candidate a is better than or equal to candidate b, and candidate b is better than or equal to candidate a, but they are not the same people, we can not decide which one is the best.
 Transitive is violated: If candidate a is better than or equal to candidate b, and candidate b is better than or equal to candidate c, but a is not better than or equal to candidate c, we can not decide which one is best.
5.12 ★
Describe an implementation of the procedure Random(a, b) that only makes calls to Random(0, 1). What is the expected running time of your procedure, as a function of a and b?
The procedure is implemented here.
let n = ⌈lg (b  a + 1)⌉ be the number of bits need to generate.
After generating n bits of number we are in a space of \(2^n\) numbers. But the space we need to generate is of size b  a + 1. We have the possibility of p = (b  a + 1) / \(2^n\) to get a usable random number, and possibility of 1  p to generate the number again.
Let k = Θ(n) be the time needed to generate an n bit random number.
T(a, b) = p k + (1  p) (k + T(a, b)).
⇒ T(a, b) = p k + (1  p) k + (1  p) T(a, b).
⇒ T(a, b) = k + (1  p) T(a, b).
⇒ T(a, b) = k / (1  (1  p)).
⇒ T(a, b) = k / p.
⇒ T(a, b) = Θ(n) / ((b  a + 1) / \(2^n\)).
⇒ T(a, b) = Θ(n) \(2^{\left\lceil\lg (b  a + 1)\right\rceil}\) / (b  a + 1).
⇒ T(a, b) = Θ(n \(2^{\left\lceil\lg (b  a + 1)\right\rceil}\) / (b  a + 1)).
⇒ T(a, b) = Θ(n \(2^{\lg (b  a + 1)}\) / (b  a + 1)),
because \(Θ(n) = 2^n ≤ 2^{\left\lceil n\right\rceil} < 2^{n + 1} = 2 × 2^n = Θ(n)\).
⇒ T(a, b) = Θ(n (b  a + 1) / (b  a + 1)).
⇒ T(a, b) = Θ(n).
⇒ T(a, b) = Θ(⌈lg (b  a + 1)⌉).
5.13 ★
Suppose that you want to output 0 with probability 1 / 2 and 1 with probability 1 / 2. At your disposal is a procedure BiasedRandom, that outputs either 0 or 1. It outputs 1 with some probability p and 0 with probability 1  p, where 0 < p < 1, but you do not know what p is. Give an algorithm that uses BiasedRandom as a subroutine, and returns an unbiased answer, returning 0 with probability 1 / 2 and 1 with probability 1 / 2. What is the expected running time of your algorithm as a function of p?
Solution is implemented here.
Let k be the time used to generate and compare two random value, T(p) be the expected running time of our algorithm, we have:
T(p) = 2 p (1  p) k + (1  2 p (1  p)) (k + T(p))
⇒ T(p) = k / (2 p (1  p)) = Θ(1 / (p (1  p))).
5.2 Indicator random variables
5.21
In HireAssistant, assuming that the candidates are presented in a random order, what is the probability that you hire exactly one time? What is the probability that you hire exactly n times?
The probability of hiring exactly one time is 1 / n.
The probability of hiring exactly n times is 1 / n!.
5.22
In HireAssistant, assuming that the candidates are presented in a random order, what is the probability that you hire exactly twice?
The first candidate will be hired, so if there are two candidates being hired, the second candidate must be the best one, and all candidates between the first candidate and the best candidate are less good than the first candidate.
Let i be the rank of the first candidate, then there is i candidates are better than the first candidate, and n  i  1 candidates are less good than the first candidate. If there are j candidates between the first candidate and the best candidate, there are P(n  i  1, j) (n  j  2)! different situations.
So the probability is
\(\frac{∑_{i = 1}^{n  1} ∑_{j = 0}^{n  i  1} P(n  i  1, j) (n  j  2)!}{n!} = H_{n  1} / n\),
where \(H_n\) is the nth harmonic number.
5.23
Use indicator random variables to compute the expected value of the sum of n dice.
Let \(X_i\) be the indicator variable associated with the event in which the value of one dice is i.
So the expected value of one dice is:
E[X] = \(∑_{i = 1}^6 i \Pr\left\lbrace X_i = 1\right\rbrace\) = \(∑_{i = 1}^6 i / 6\) = 7 / 2.
So the expected value of n dices is:
\(\operatorname{E}\left[∑_{j = 1}^{n} X\right]\) = \(∑_{j = 1}^{n}\operatorname{E}[X]\) = 7 n / 2.
5.24
Use indicator random variables to solve the following problem, which is known as the hatcheck problem. Each of n customers gives a hat to a hatcheck person at a restaurant. The hatcheck person gives the hats back to the customers in a random order. What is the expected number of customers who get back their own hat?
Let \(X_i\) be the indicator variable associated with the event in which the ith customer get back its own hat.
E[X]
= \(\operatorname{E}\left[∑_{i = 1}^n X_i\right]\)
= \(∑_{i = 1}^n \operatorname{E}\left[X_i\right]\)
= \(∑_{i = 1}^n 1 / n\)
= \(∑_{i = 1}^n 1 / n\)
= 1.
5.25
Let A[1‥n] be an array of n distinct numbers. If i < j and A[i] > A[j], then the pair (i, j) is called an inversion of A. (See Problem 24 for more on inversions.) Suppose that the elements of A form a uniform random permutation of ⟨1, 2, …, n⟩. Use indicator random variables to compute the expected number of inversions.
Let \(X_{i j}\) be the indicator variable associated with the event in which the A[i] > A[j], where i < j.
E[X]
= \(E\left[∑_{i = 1}^{n  1} ∑_{j = i + 1}^{n} X_{i j}\right]\)
= \(∑_{i = 1}^{n  1} E\left[∑_{j = i + 1}^{n} X_{i j}\right]\)
= \(∑_{i = 1}^{n  1} ∑_{j = i + 1}^{n} E\left[X_{i j}\right]\)
= \(∑_{i = 1}^{n  1} ∑_{j = i + 1}^{n} 1 / 2\)
= \(∑_{i = 1}^{n  1} (n  i) / 2\)
= n (n  1) / 4.
5.31
Professor Marceau objects to the loop invariant used in the proof of Lemma 5.5. He questions whether it is true prior to the first iteration. He reasons that we could just as easily declare that an empty subarray contains no 0permutations. Therefore, the probability that an empty subarray contains a 0permutation should be 0, thus invalidating the loop invariant prior to the first iteration. Rewrite the procedure RandomizeInPlace so that its associated loop invariant applies to a nonempty subarray prior to the first iteration, and modify the proof of Lemma 5.5 for your procedure.
Solution is implemented here.
5.32
Professor Kelp decides to write a procedure that produces at random any permutation besides the identity permutation. He proposes the following procedure:
PermuteWithoutIdentity(A)
 n = A.length
 for i = 1 to n  1
 swap A[i] with A[Random(i + 1, n)]
Does this code do what Professor Kelp intends?
No, the algorithm above could not produce the permutation ⟨A[1], A[3], A[2], A[4], A[5], …, A[n]⟩ which is not the identity permutation, because the first element always is always swapped away from its original position.
5.33
Suppose that instead of swapping element A[i] with a random element from the subarray A[i‥n], we swapped it with a random element from anywhere in the array:
PermuteWithAll(A)
 n = A.length
 for i = 1 to n
 swap A[i] with A[Random(1, n)]
Does this code produce a uniform random permutation? Why or why not?
In each iteration, there are n different possible outcomes of the Random function with the same possibility, so there are total of \(n^n\) (possibly same) outcomes of the algorithm with the same possibility. But the uniform random permutation requires there are n! different possible outcomes. It is possible that \(n^n / n!\) is not an integer, in which case, it is impossible to divide \(n^n\) probabilities into n! same probabilities. So, the code does not produce a uniform random permutation.
5.34
Professor Armstrong suggests the following procedure for generating a uniform random permutation:
PermuteByCyclic(A)
 n = A.length
 let B[1‥n] be a new array
 offset = Random(1, n)
 for i = 1 to n
 dest = i + offset
 if dest > n
 dest = dest  n
 B[dest] = A[i]
 return B
Show that each element A[i] has a 1 / n probability of winding up in any particular position in B. Then show that Professor Armstrong is mistaken by showing that the resulting permutation is not uniformly random.
The permutation is generated by shifting the array by length Random(1, n). So for each position, any element can wind up at that position with probability of 1 / n.
There can only be n different outcomes with PermuteByCyclic, while the uniform distribution requires n! different results. So the result permutation can not be uniformly random.
5.35 ★
Prove that in the array P in procedure PermuteBySorting, the probability that all elements are unique is at least 1  1 / n.
The probability is:
\(P\left(n^3, n\right) / \left(n^3\right)^n\)
= \(\left(n^3! / \left(n^3  n\right)!\right) / \left(n^3\right)^n\)
= \(\left(∏_{i = 0}^{n  1} \left(n^3  i\right)\right) / \left(∏_{i = 0}^{n  1} n^3\right)\)
= \(∏_{i = 0}^{n  1} \left(\left(n^3  i\right) / n^3\right)\)
= \(∏_{i = 0}^{n  1} \left(1  i / n^3\right)\)
≥ \(∏_{i = 0}^{n  1} \left(1  n / n^3\right)\)
= \(∏_{i = 0}^{n  1} \left(1  1 / n^2\right)\)
= \(\left(1  1 / n^2\right)^n\)
Not sure how to go from here.
5.36
Explain how to implement the algorithm PermuteBySorting to handle the case in which two or more priorities are identical. That is, your algorithm should produce a uniform random permutation, even if two or more priorities are identical.
Skipped.
5.37
Suppose we want to create a random sample of the set {1, 2, 3, …, n}, that is, an melement subset S, where 0 ≤ m ≤ n, such that each msubset is equally likely to be created. One way would be to set A[i] = i for i = 1, 2, 3, …, n, call RandomizeInPlace(A), and then take just the first m array elements. This method would make n calls to the Random procedure. If n is much larger than m, we can create a random sample with fewer calls to Random. Show that the following recursive procedure returns a random msubset S of {1, 2, 3, …, n}, in which each msubset is equally likely, while making only m calls to Random:
RandomSample(m, n)
 if m == 0
 return ∅
 else S = RandomSample(m  1, n  1)
 i = Random(1, n)
 if i ∈ S
 S = S ∪ {n}
 else S = S ∪ {i}
 return S
The algorithm is implemented here.
Like exercise 5.34, it is not enough by just proving ∀ i ∈ ℕ and 1 ≤ i ≤ n, the probability of i ∈ RandomSample(m, n) is m / n.
Lemma: RandomSample(m, n) generates any msubset with the probability of 1 / C(n, m).
Proof by induction on m:

Base case: RandomSample(0, n) only returns the empty set, and the only 0subset is the empty set, so the probability is 1, which equals to 1 / C(n, 0), the claim holds.

Inductive cases:

For a certain msubset containing n, there are two different ways to generate it:
 When i ∈ S, which has the probability of (m  1) / n,
 Or i = n, which has the probability of 1 / n,
So the probability of generating an msubset containing n is (m  1) / n + 1 / n = m / n.
By induction, generating an (m  1)subset on the set [1, n  1] has the probability of 1 / C(n  1, m  1), so generating an msubset on the set [1, n] on set [1, n  1] has the probability of (1 / C(n  1, m  1)) (m / n) = 1 / C(n, m).

For a certain msubset does not contain n, let the last element being added to S can be k, then k can be any element in the msubset.
So the probability of generating a certain msubset does not contain n is \(∑_{j = 1}^{m} (1 / C(n  1, m  1)) (1 / n)\) = (1 / C(n  1, m  1)) (m / n) = 1 / C(n, m).
So all msubsets have the probability of 1 / C(n, m) of being generated, the claim holds.

5.4 Probabilistic analysis and further uses of indicator random variables ★
5.41
How many people must there be in a room before the probability that someone has the same birthday as you do is at least 1 / 2? How many people must there be before the probability that at least two people have a birthday on July 4 is greater than 1 / 2?

The probability that no people in a room has the same birthday as me is \(364^{k  1} / 365^{k  1}\), so the probability that someone has the same birthday as me is \(1  364^{k  1} / 365^{k  1}\).
Solving 1  \(364^{k  1} / 365^{k  1}\) ≥ 1 / 2, we get k ≥ 254.

The probability that x people in a room has the a birthday on July 4 is C(k, x) \(364^{k  x} / 365^k\),
Solving 1  C(k, 0) \(364^k / 365^k\)  C(k, 1) \(364^{k  1} / 365^k\) > 1 / 2, we get k ≥ 613.
5.42
Suppose that we toss balls into b bins until some bin contains two balls. Each toss is independent, and each ball is equally likely to end up in any bin. What is the expected number of ball tosses?
The probability that toss k balls into b bins to get the first bin containing two balls is: (P(b, k  1) / \(b^{k  1}\)) ((k  1) / b).
So the expected number of ball tosses is:
\(∑_{k = 2}^{b + 1} \left(P(b, k  1) / b^{k  1}\right) ((k  1) / b) k\)
5.43 ★
For the analysis of the birthday paradox, is it important that the birthdays be mutually independent, or is pairwise independence sufficient? Justify your answer.
Skipped.
5.44 ★
How many people should be invited to a party in order to make it likely that there are three people with the same birthday?
Skipped.
5.45 ★
What is the probability that a kstring over a set of size n forms a kpermutation? How does this question relate to the birthday paradox?
The probability is P(n, k) / \(n^k\).
It’s the same as saying k people whose birthdays are in n days all have different birthday.
5.46 ★
Suppose that n balls are tossed into n bins, where each toss is independent and the ball is equally likely to end up in any bin. What is the expected number of empty bins? What is the expected number of bins with exactly one ball?
Let \(X_{i}\) being the indicator variable in which bin i have k balls, then
E[\(X_{i k}\)] = C(n, k) \((1 / n)^k (1  1 / n)^{n  k}\).
So the expected number of bins with exactly k balls is
E[\(∑_{i = 1}^{n} X_{i k}\)]
= \(∑_{i = 1}^{n} \operatorname{E}\left[X_{i k}\right]\)
= \(∑_{i = 1}^{n} C(n, k) (1 / n)^k (1  1 / n)^{n  k}\)
= n C(n, k) \((1 / n)^k (1  1 / n)^{n  k}\).
So the expected number of bins with exactly 0 ball is:
n C(n, 0) \((1 / n)^0 (1  1 / n)^{n  0}\)
= n \((1  1 / n)^n\).
So the expected number of bins with exactly 1 ball is:
n C(n, 1) \((1 / n)^1 (1  1 / n)^{n  1}\)
= n \((1  1 / n)^{n  1}\).
5.47 ★
Sharpen the lower bound on streak length by showing that in n flips of a fair coin, the probability is less than 1 / n that no streak longer than lg n  2 lg lg n consecutive heads occurs.
Skipped.
5.X Problems
51 Probabilistic counting
With a bbit counter, we can ordinarily only count up to \(2^b\)  1. With R. Morris’s probabilistic counting, we can count up to a much larger value at the expense of some loss of precision.
We let a counter value of i represent a count of \(n_i\) for i = 0, 1, …, \(2^b\)  1, where the \(n_i\) form an increasing sequence of nonnegative values. We assume that the initial value of the counter is 0, representing a count of \(n_0\) = 0. The Increment operation works on a counter containing the value i in a probabilistic manner. If i = \(2^b\)  1, then the operation reports an overflow error. Otherwise, the Increment operation increases the counter by 1 with probability 1 / (\(n_{i + 1}  n_i\)), and it leaves the counter unchanged with probability 1  1 / (\(n_{i + 1}  n_i\)).
If we select \(n_i\) = i for all i ≥ 0, then the counter is an ordinary one. More interesting situations arise if we select, say, \(n_i = 2^{i  1}\) for i > 0 or \(n_i = F_i\) (the ith Fibonacci number—see Section 3.2).
For this problem, assume that \(n_{2^b  1}\) is large enough that the probability of an overflow error is negligible.
 Show that the expected value represented by the counter after n Increment operations have been performed is exactly n.
 The analysis of the variance of the count represented by the counter depends on the sequence of the \(n_i\). Let us consider a simple case: \(n_i\) = 100 i for all i ≥ 0. Estimate the variance in the value represented by the register after n Increment operations have been performed.

For any counter value \(n_i\), the expected value increased by one increment operation is:
\((n_{i + 1}  n_i) (1 / (n_{i + 1}  n_i)) + 0 (1 / (n_{i + 1}  n_i))\) = 1.
Let \(X_i\) be the random variable in which the counter values increased by the ith increment operation. So the expected value of the counter after n Increment operations is:
E[\(∑_{i = 1}^{n} X_i\)]
= \(∑_{i = 1}^{n} \operatorname{E}\left[X_i\right]\)
= \(∑_{i = 1}^{n} 1\).
= n. 
Skipped.
52 Searching an unsorted array
This problem examines three algorithms for searching for a value x in an unsorted array A consisting of n elements.
Consider the following randomized strategy: pick a random index i into A. If A[i] = x, then we terminate; otherwise, we continue the search by picking a new random index into A. We continue picking random indices into A until we find an index j such that A[j] = x or until we have checked every element of A. Note that we pick from the whole set of indices each time, so that we may examine a given element more than once.
a. Write pseudocode for a procedure RandomSearch to implement the strategy above. Be sure that your algorithm terminates when all indices into A have been picked.
b. Suppose that there is exactly one index i such that A[i] = x. What is the expected number of indices into A that we must pick before we find x and RandomSearch terminates?
c. Generalizing your solution to part (b), suppose that there are k ≥ 1 indices i such that A[i] = x. What is the expected number of indices into A that we must pick before we find x and RandomSearch terminates? Your answer should be a function of n and k.
d. Suppose that there are no indices i such that A[i] = x. What is the expected number of indices into A that we must pick before we have checked all elements of A and RandomSearch terminates?
Now consider a deterministic linear search algorithm, which we refer to as DeterministicSearch. Specifically, the algorithm searches A for x in order, considering A[1], A[2], A[3], …, A[n] until either it finds A[i] = x or it reaches the end of the array. Assume that all possible permutations of the input array are equally likely.
e. Suppose that there is exactly one index i such that A[i] = x. What is the averagecase running time of DeterministicSearch? What is the worstcase running time of DeterministicSearch?
f. Generalizing your solution to part (e), suppose that there are k ≥ 1 indices i such that A[i] = x. What is the averagecase running time of DeterministicSearch? What is the worstcase running time of DeterministicSearch? Your answer should be a function of n and k.
g. Suppose that there are no indices i such that A[i] = x. What is the averagecase running time of DeterministicSearch? What is the worstcase running time of DeterministicSearch?
Finally, consider a randomized algorithm ScrambleSearch that works by first randomly permuting the input array and then running the deterministic linear search given above on the resulting permuted array.
h. Letting k be the number of indices i such that A[i] = x, give the worstcase and expected running times of ScrambleSearch for the cases in which k = 0 and k = 1. Generalize your solution to handle the case in which k ≥ 1.
i. Which of the three searching algorithms would you use? Explain your answer.
a.
ScrambleSearch(A, v)
 V = an array with A.length False values
 c = 0
 while c < A.length
 i = Random(1, n)
 if not V[i]
 if A[i] == v
 return i
 V[i] = True
 c = c + 1
 return nil
b. Each pick has the probability of 1 / n of being succeed, so the expected number of indices to pick is n.
c. Each pick has the probability of k / n of being succeed, so the expected number of indices to pick is n / k.
d. It is like the coupon collector’s problem, the expected number of indices to pick is n(ln n + O(1)).
e. Averagecase running time is Θ((1 + n) / 2), Worstcase running time is Θ(n).
f. The expected number of indices to pick is:
\(∑_{i = 1}^{n  k + 1} i P(n  k, i  1) k P(n  i, n  i) / P(n, n)\) = (n + 1) / (k + 1). (By WolframAlpha)
Worstcase running time is Θ(n  k + 1).
g. They are both Θ(n).
h. The worstcase, expected running time are all Θ(n).
i. I will use DeterministicSearch, it is superior in every way.
II Sorting and Order Statistics
6 Heapsort
6.1 Heaps
6.11
What are the minimum and maximum numbers of elements in a heap of height h?
The minimum number of elements is \(2^h\), the maximum number of elements is \(2^{h + 1}  1\).
6.12
Show that an nelement heap has height ⌊lg n⌋.
According to exercise 6.11, we know that if the height of a heap is h, the number of elements is in range [\(2^h\), \(2^{h + 1}  1\)]. Also, if \(2^h\) ≤ n ≤ \(2^{h + 1}  1\), we must have ⌊lg n⌋ = h, so an nelement heap must have height ⌊lg n⌋.
6.13
Show that in any subtree of a maxheap, the root of the subtree contains the largest value occurring anywhere in that subtree.
Proof by induction:
Base case: if a maxheap has one element, then the only subtree is the oneelement tree, the root contains the largest value, since it is the only value the tree has.
Inductive case: if a maxheap H has more than one element, by induction, we know that for any subtree in the H’s children, the subtree’s root contains the largest value in the subtree. So the left child H contains the largest value in the left subtree; and if H have a right child, it must also contain the largest value in the right subtree. Since we have that for every node in the heap, the value of the parent is always greater or equal than its children, we know that the value of the root of H must be always greater or equal than its children. So it must be the largest element in the heap.
6.14
Where in a maxheap might the smallest element reside, assuming that all elements are distinct?
The smallest element must be in the leaf nodes, otherwise the heap property will be violated.
6.15
Is an array that is in sorted order a minheap?
Yes, since the index of a child node must be greater than the corresponding parent node, and the array is in sorted order, we know that the value of the child node must be greater or equal to the corresponding parent node, which satisfies the minheap property.
6.16
Is the array with values ⟨23, 17, 14, 6, 13, 10, 1, 5, 7, 12⟩ a maxheap?
Nope.
6.17
Show that, with the array representation for storing an nelement heap, the leaves are the nodes indexed by ⌊n / 2⌋ + 1, ⌊n / 2⌋ + 2, …, n.
The index of the last element is n, so the index of the parent of the last element is ⌊n / 2⌋, so the leaves are the nodes starts with ⌊n / 2⌋, and ends with n.
6.2 Maintaining the heap property
6.21
Using Figure 6.2 as a model, illustrate the operation of MaxHeapify(A, 3) on the array A = ⟨27, 17, 3, 16, 13, 10, 1, 5, 7, 12, 4, 8, 9, 0⟩.
Skipped.
6.22
Starting with the procedure MaxHeapify, write pseudocode for the procedure MinHeapify(A, i), which performs the corresponding manipulation on a minheap. How does the running time of MinHeapify compare to that of MaxHeapify?
Solution is implemented here.
The running time of MinHeapify is the same as MaxHeapify.
6.23
What is the effect of calling MaxHeapify(A, i) when the element A[i] is larger than its children?
Nothing happens.
6.24
What is the effect of calling MaxHeapify(A, i) for i > A.heapsize / 2?
Nothing happens.
6.25
The code for MaxHeapify is quite efficient in terms of constant factors, except possibly for the recursive call in line 10, which might cause some compilers to produce inefficient code. Write an efficient MaxHeapify that uses an iterative control construct (a loop) instead of recursion.
The solution is implemented here.
6.26
Show that the worstcase running time of MaxHeapify on a heap of size n is Ω(lg n). (Hint: For a heap with n nodes, give node values that cause MaxHeapify to be called recursively at every node on a simple path from the root down to a leaf.)
The height of a heap of size n is ⌊lg n⌋, if MaxHeapify is called at every node on a simple path from the root down to a leaf, the expected running time is linear to the height of the heap, that is ⌊lg n⌋, so the worstcase running time is Ω(lg n).
6.3 Building a heap
6.31
Using Figure 6.3 as a model, illustrate the operation of BuildMaxHeap on the array A = ⟨5, 3, 17, 10, 84, 19, 6, 22, 9⟩.
Skipped.
6.32
Why do we want the loop index i in line 2 of BuildMaxHeap to decrease from ⌊A.length / 2⌋ to 1 rather than increase from 1 to ⌊A.length / 2⌋?
Because MaxHeapify requires that its child subtree being a heap, so we have to do this bottomup style.
6.33
Show that there are at most ⌈n / \(2^{h + 1}\)⌉ nodes of height h in any nelement heap.
Proof by induction:
Base case: the nodes of height 0 are leaf nodes. according to exercise 6.17, we know that the number of leaf nodes is n  (⌊n / 2⌋ + 1) + 1 = n  ⌊n / 2⌋ = ⌈n / 2⌉ = ⌈n / \(2^{0 + 1}\)⌉, since “For any integer n, ⌈n / 2⌉ + ⌊n / 2⌋ = n”.
Inductive case: by induction, we know that in height h + 1, there are at most ⌈n / \(2^{h + 2}\)⌉ nodes. So there are at most ⌈⌈n / \(2^{h + 2}\)⌉ / 2⌉ nodes in height h. According to equation (3.4), we have ⌈⌈n / \(2^{h + 2}\)⌉ / 2⌉ = ⌈n / \(2^{h + 2}\) / 2⌉ = ⌈n / \(2^{h + 1}\)⌉.
6.4 The heapsort algorithm
6.41
Using Figure 6.4 as a model, illustrate the operation of Heapsort on the array A = ⟨5, 13, 2, 25, 7, 17, 20, 8, 4⟩.
Skipped.
6.42
Argue the correctness of Heapsort using the following loop invariant:
At the start of each iteration of the for loop of lines 2–5, the subarray A[1‥i] is a maxheap containing the i smallest elements of A[1‥n], and the subarray A[i + 1‥n] contains the n  i largest elements of A[1‥n], sorted.
Skipped.
6.43
What is the running time of Heapsort on an array A of length n that is already sorted in increasing order? What about decreasing order?
Both Θ(n lg n).
6.44
Show that the worstcase running time of Heapsort is Ω(n lg n).
Skipped.
6.45 ★
Show that when all elements are distinct, the bestcase running time of Heapsort is Ω(n lg n).
Skipped.
6.5 Priority queues
6.51
Illustrate the operation of HeapExtractMax on the heap A = ⟨15, 13, 9, 5, 12, 8, 7, 4, 0, 6, 2, 1⟩.
Skipped.
6.52
Illustrate the operation of MaxHeapInsert(A, 10) on the heap A = ⟨15, 13, 9, 5, 12, 8, 7, 4, 0, 6, 2, 1⟩.
Skipped.
6.53
Write pseudocode for the procedures HeapMinimum, HeapExtractMin, HeapDecreaseKey, and MinHeapInsert that implement a minpriority queue with a minheap.
Solution is implemented here.
6.54
Why do we bother setting the key of the inserted node to ∞ in line 2 of MaxHeapInsert when the next thing we do is increase its key to the desired value?
We must ensure that when calling HeapIncreaseKey, the value of key must be equal to or greater than A[i], so we use ∞ to do this.
6.55
Argue the correctness of HeapIncreaseKey using the following loop invariant:
At the start of each iteration of the while loop of lines 4–6, A[Parent(i)] ≥ A[Left(i)] and A[Parent(i)] ≥ A[Right(i)], if these nodes exist, and the subarray A[1‥A.heapsize] satisfies the maxheap property, except that there may be one violation: A[i] may be larger than A[Parent(i)].
You may assume that the subarray A[1‥A.heapsize] satisfies the maxheap property at the time HeapIncreaseKey is called.
Skipped.
6.56
Each exchange operation on line 5 of HeapIncreaseKey typically requires three assignments. Show how to use the idea of the inner loop of InsertionSort to reduce the three assignments down to just one assignment.
Solution is implemented here.
6.57
Show how to implement a firstin, firstout queue with a priority queue. Show how to implement a stack with a priority queue. (Queues and stacks are defined in Section 10.1.)
Solution is implemented here.
6.58
The operation HeapDelete(A, i) deletes the item in node i from heap A. Give an implementation of HeapDelete that runs in O(lg n) time for an nelement maxheap.
Solution is implemented here.
6.59
Give an O(n lg k)time algorithm to merge k sorted lists into one sorted list, where n is the total number of elements in all the input lists. (Hint: Use a minheap for kway merging.)
Solution is implemented here.
6.X Problems
61 Building a heap using insertion
We can build a heap by repeatedly calling MaxHeapInsert to insert the elements into the heap. Consider the following variation on the BuildMaxHeap procedure:
BuildMaxHeap′(A)
 A.heapsize = 1
 for i = 2 to A.length
 BuildMaxHeap′(A, A[i])
a. Do the procedures BuildMaxHeap and BuildMaxHeap′ always create the same heap when run on the same input array? Prove that they do, or provide a counterexample.
b. Show that in the worst case, BuildMaxHeap′ requires Θ(n lg n) time to build an nelement heap.
a. No. For array ⟨0, 1, 2, 3⟩, BuildMaxHeap produces ⟨3, 1, 2, 0⟩, while BuildMaxHeap′ produces ⟨3, 2, 1, 0⟩.
b. If A is in increasing order, every insertion will have to travel all the way from leaf to root, which has the run time of Θ(n lg n).
62 Analysis of dary heaps
A dary heap is like a binary heap, but (with one possible exception) nonleaf nodes have d children instead of 2 children.
a. How would you represent a dary heap in an array?
b. What is the height of a dary heap of n elements in terms of n and d?
c. Give an efficient implementation of ExtractMax in a dary maxheap. Analyze its running time in terms of d and n.
d. Give an efficient implementation of Insert in a dary maxheap. Analyze its running time in terms of d and n.
e. Give an efficient implementation of IncreaseKey(A, i, k), which flags an error if k < A[i], but otherwise sets A[i] = k and then updates the dary maxheap structure appropriately. Analyze its running time in terms of d and n.
a. The solution is implemented here.
b. The maximum number of elements of height h is \(∑_{i = 0}^{h} d^h = \frac{1  d^h}{1  d}\), so we have \(\frac{1  d^{h  1}}{1  d} < n ≤ \frac{1  d^h}{1  d}\). Solving it, we have \(\log_d((d  1) n + 1) ≤ h < \log_d((d  1) n + 1) + 1\). Since h is an integer, we have h = \(\left\lceil\log_d((d  1) n + 1)\right\rceil\) = Θ(\(\log_d n\)).
c. The solution is implemented here. The worstcase running time is linear to the height of the heap, and on every level we have to find the maximum element in Θ(d) time. So the worst case running time is Θ(d \(\log_d n\)).
d. The solution is implemented here. The worstcase running time is linear to the height of the heap. So the worst case running time is O(\(\log_d n\)).
e. The solution is implemented here. The worst case running time is Θ(\(\log_d n\)).
63 Young tableaus
An m × n Young tableau is an m × n matrix such that the entries of each row are in sorted order from left to right and the entries of each column are in sorted order from top to bottom. Some of the entries of a Young tableau may be ∞, which we treat as nonexistent elements. Thus, a Young tableau can be used to hold r ≤ m n finite numbers.
a. Draw a 4 × 4 Young tableau containing the elements {9, 16, 3, 2, 4, 8, 5, 14, 12}.
b. Argue that an m × n Young tableau Y is empty if Y[1, 1] = ∞. Argue that Y is full (contains m n elements) if Y[m, n] < ∞.
c. Give an algorithm to implement ExtractMin on a nonempty m × n Young tableau that runs in O(m + n) time. Your algorithm should use a recursive subroutine that solves an m × n problem by recursively solving either an (m  1) × n or an m × (n  1) subproblem. (Hint: Think about MaxHeapify.) Define T(p), where p = m + n, to be the maximum running time of ExtractMin on any m × n Young tableau. Give and solve a recurrence for T(p) that yields the O(m + n) time bound.
d. Show how to insert a new element into a nonfull m × n Young tableau in O(m + n) time.
e. Using no other sorting method as a subroutine, show how to use an n × n Young tableau to sort \(n^2\) numbers in O(\(n^3\)) time.
f. Give an O(m + n)time algorithm to determine whether a given number is stored in a given m × n Young tableau.
a.
2  3  4  5 
8  9  12  14 
16  ∞  ∞  ∞ 
∞  ∞  ∞  ∞ 
b.
 Since Y[1, 1] = ∞, the first row will only contain ∞, which means that the first element of every column is ∞, which means every column will only contain ∞, which means the table will only contain ∞, which means Y is empty.
 Since Y[m, n] < ∞, the last row will not contain ∞, which means that the last element of every column is not ∞, which means every column will not contain ∞, which means the table will not contain ∞, which means Y full.
c.
Solution is implemented here.
T(p) = T(p  1) + O(1).
d.
Solution is implemented here.
The solution is very similar to the ExtractMin and they have the same running time bound.
e.
Solution is implemented here.
f.
You are thinking about binary search, aren’t you? Too bad this problem isn’t about that.
Solution is implemented here.
7 Quicksort
7.1 Description of quicksort
7.11
Using Figure 7.1 as a model, illustrate the operation of Partition on the array A = ⟨13, 19, 9, 5, 12, 8, 7, 4, 21, 2, 6, 11⟩.
Skipped.
7.12
What value of q does Partition return when all elements in the array A[p‥r] have the same value? Modify Partition so that q = ⌊(p + r) / 2⌋ when all elements in the array A[p‥r] have the same value.
 Partition returns r when all elements have the same value.
 Solution is implemented here.
7.13
Give a brief argument that the running time of Partition on a subarray of size n is Θ(n).
There is only one loop that updates j from p to r  1, and the loop body requires constant time to run, so the running time is Θ(r  p) = Θ(n  1) = Θ(n).
7.14
How would you modify Quicksort to sort into nonincreasing order?
Do you mean that I can’t modify Partition?
Fine, the solution is implemented here.
7.2 Performance of quicksort
7.21
Use the substitution method to prove that the recurrence T(n) = T(n  1) + Θ(n) has the solution T(n) = Θ(\(n^2\)), as claimed at the beginning of Section 7.2.
Let T(n) = T(n  1) + f(n), where for some \(c_1\), \(c_2\) and \(n_0\), for any n > \(n_0\), \(c_1 n ≤ f(n) ≤ c_2 n\).
Suppose T(n) ≥ \(\left(c_1 / 2\right) n^2 + \left(c_1 / 2\right) n + c_3\) for large enough n, we have
T(n)
= T(n  1) + f(n)
≥ T(n  1) + \(c_1\) n
≥ \(\left(c_1 / 2\right) (n  1)^2 + \left(c_1 / 2\right) (n  1) + c_3 + c_1 n\)
= \(\left(c_1 / 2\right) n^2  c_1 n + c_1 / 2 + \left(c_1 / 2\right) (n  1) + c_3 + c_1 n\)
= \(\left(c_1 / 2\right) n^2 + \left(c_1 / 2\right) n + c_3\).
So T(n) = Ω(\(n^2\)). Similarly, we can prove T(n) = O(\(n^2\)), so T(n) = Θ(n).
7.22
What is the running time of Quicksort when all elements of array A have the same value?
Θ(\(n^2\)).
7.23
Show that the running time of Quicksort is Θ(\(n^2\)) when the array A contains distinct elements and is sorted in decreasing order.
Let’s say we have an array ⟨n, n  1, …, 2, 1⟩ to sort.
At the first partition, we will choose 1 as the pivot element. so after partition, we get two arrays: ⟨⟩ and ⟨n  1, n  2, …, 2, n⟩. This is the worst case scenario.
At the second partition, we will choose n as the pivot element. so after partition, we get two arrays: ⟨n  1, n  2, …, 2⟩ and ⟨⟩. This is the worst case scenario.
Notice that after the second partition, the subarray ⟨n  1, n  2, …, 2⟩ is in decreasing order, so we get ourselves a subproblem of the original problem. Since the first two partitions of the partition is the worst cast scenario, we know that the subproblem will also be the worst case scenario. So the running time is Θ(\(n^2\)).
7.24
Banks often record transactions on an account in order of the times of the transactions, but many people like to receive their bank statements with checks listed in order by check number. People usually write checks in order by check number, and merchants usually cash them with reasonable dispatch. The problem of converting timeoftransaction ordering to checknumber ordering is therefore the problem of sorting almostsorted input. Argue that the procedure InsertionSort would tend to beat the procedure Quicksort on this problem.
According to problem 24, the InsertionSort algorithm has the running time of Θ(k), where k is the inversions of the array. An almost sorted array as a few number of inversions, so the expected running time of InsertionSort should be low, while the Quicksort algorithm requires at Ω(n lg n) time. So InsertionSort would tend to beat the procedure Quicksort on this problem.
7.25
Suppose that the splits at every level of quicksort are in the proportion 1  α to α, where 0 < α ≤ 1 / 2 is a constant. Show that the minimum depth of a leaf in the recursion tree is approximately lg n / lg α and the maximum depth is approximately lg n / lg(1  α). (Don’t worry about integer roundoff.)
At each level of recursion the problem of size n is divided into problems of size α n and size (1  α) n. Since α ≤ 1 / 2, the minimum depth is on the path n → α n → \(\alpha^2\) n → ⋯ → 1, the maximum depth is on the path n → (1  α) n → \((1  \alpha)^2\) n → ⋯ → 1.
Solving \(\alpha^i\) n = 1 for i, we get i = lg n / lg α.
Solving \((1  \alpha)^i\) n = 1 for i, we get i = lg n / lg (1  α).
7.26 ★
Argue that for any constant 0 < α ≤ 1 / 2, the probability is approximately 1  2 α that on a random input array, Partition produces a split more balanced than 1  α to α.
Let n be the size of the array. Being more balanced than 1  α to α means that the number of elements is greater than or equal to A[n] is greater than a n and less than (1  a) n, which means A[n] must be greater than the (a n)th smallest element and less than the ((1  a) n)th biggest element. There are (1  a) n  a n = (1  2 a) n different cases to choose. So the probability is (1  2 a) n / n = 1  2 a.
7.3 A randomized version of quicksort
7.31
Why do we analyze the expected running time of a randomized algorithm and not its worstcase running time?
The worstcase running time doesn’t change if we randomize the algorithm, since we might generate the very same worst case by chance.
7.32
When RandomizedQuicksort runs, how many calls are made to the randomnumber generator Random in the worst case? How about in the best case? Give your answer in terms of Θnotation.
 Best case: T(n) = 2 T((n  1) / 2) + 1 ⇒ T(n) = Θ(n).
 Worst case: T(n) = T(n  1) + 1 ⇒ T(n) = Θ(n).
7.4 Analysis of quicksort
7.41
Show that in the recurrence
T(n) = \(\underset{0 ≤ q ≤ n  1}{\max} (T(q) + T(n  q  1)) + Θ(n)\),
T(n) = Ω(\(n^2\)).
Suppose T(n) ≥ c \(n^2\) for some c, we have:
T(n)
= \(\underset{0 ≤ q ≤ n  1}{\max} (T(q) + T(n  q  1)) + Θ(n)\)
≥ \(\underset{0 ≤ q ≤ n  1}{\max} \left(c q^2 + c (n  q  1)^2\right)\) + Θ(n)
= c ⋅ \(\underset{0 ≤ q ≤ n  1}{\max} \left(q^2 + (n  q  1)^2\right)\) + Θ(n)
= c ⋅ \((n  1)^2\) + Θ(n)
= c \(n^2\)  c (2 n  1) + Θ(n)
≥ c \(n^2\).
7.42
Show that quicksort’s bestcase running time is Ω(n lg n).
The best case running time is:
T(n) = 2 T((n  1) / 2) + Θ(n).
Suppose T(n) ≥ c (n + 1) lg (n + 1) for some c, we have:
T(n)
= 2 T((n  1) / 2) + Θ(n)
≥ 2 c ((n  1) / 2 + 1) lg ((n  1) / 2 + 1) + Θ(n)
= 2 c ((n + 1) / 2) lg ((n + 1) / 2) + Θ(n)
= c (n + 1) (lg (n + 1)  1) + Θ(n)
= c (n + 1) lg (n + 1)  c (n + 1) + Θ(n)
≥ c (n + 1) lg (n + 1).
7.43
Show that the expression \(q^2 + (n  q  1)^2\) achieves a maximum over q = 0, 1, …, n  1 when q = 0 or q = n  1.
\(\frac{\partial\left(q^2 + (n  q  1)^2\right)}{\partial q}\)
= 2 q  2 (n  q  1)
= 4 q  2 n + 2
When q < (n  1) / 2, 4 q  2 n + 2 > 0; when q > (n  1) / 2, 4 q  2 n + 2 < 0, so we have the value of \(q^2 + (n  q  1)^2\) decreasing when q < (n  1) / 2, and increasing when q > (n  1) / 2. Since 0 ≤ q ≤ n  1, we have maximum when q = 0 or q = n  1.
7.44
Show that RandomizedQuicksort’s expected running time is Ω(n lg n).
Skipped.
7.45
We can improve the running time of quicksort in practice by taking advantage of the fast running time of insertion sort when its input is “nearly” sorted. Upon calling quicksort on a subarray with fewer than k elements, let it simply return without sorting the subarray. After the toplevel call to quicksort returns, run insertion sort on the entire array to finish the sorting process. Argue that this sorting algorithm runs in O(n k + n lg (n / k)) expected time. How should we pick k, both in theory and in practice?
Skipped.
7.46 ★
Consider modifying the Partition procedure by randomly picking three elements from array A and partitioning about their median (the middle value of the three elements). Approximate the probability of getting at worst an αto(1  α) split, as a function of α in the range 0 < α < 1.
Skipped.
7.X Problems
71 Hoare partition correctness
The version of Partition given in this chapter is not the original partitioning algorithm. Here is the original partition algorithm, which is due to C. A. R. Hoare:
HoarePartition(A, p, r)
 x = A[p]
 i = p  1
 j = r + 1
 while True
 repeat
 j = j  1
 until A[j] ≤ x
 repeat
 i = i + 1
 until A[i] ≥ x
 if i < j
 exchange A[i] with A[j]
 else return j
a. Demonstrate the operation of HoarePartition on the array A = ⟨13, 19, 9, 5, 12, 8, 7, 4, 11, 2, 6, 21⟩, showing the values of the array and auxiliary values after each iteration of the while loop in lines 4–13.
The next three questions ask you to give a careful argument that the procedure HoarePartition is correct. Assuming that the subarray A[p‥r] contains at least two elements, prove the following:
b. The indices i and j are such that we never access an element of A outside the subarray A[p‥r].
c. When HoarePartition terminates, it returns a value j such that p ≤ j < r.
d. Every element of A[p‥j] is less than or equal to every element of A[j + 1‥r] when HoarePartition terminates.
The Partition procedure in Section 7.1 separates the pivot value (originally in A[r]) from the two partitions it forms. The HoarePartition procedure, on the other hand, always places the pivot value (originally in A[p]) into one of the two partitions A[p‥j] and A[j + 1‥r]. Since p ≤ j < r, this split is always nontrivial.
e. Rewrite the Quicksort procedure to use HoarePartition.
a.
Skipped.
b.
At the first iteration of the outer loop, the algorithm first find the first element from right that is less than or equal to x, and the index element won’t be less than p since A[p] = x. Then the program find the first element that is greater or equal than x, which will have the index of p. If the algorithm doesn’t terminate, the rest of the first element is just swapping element A[i] and A[j]. So the first iteration won’t access outside the subarray A[p‥r].
After the first iteration, the following loop invariant will hold for each outer loop iteration:
 At the beginning of each iteration, i < j, and A[i] ≤ x, A[j] ≥ x.
 In the body of each iteration, i will be less than or equal to the initial value of j, and j will be more than or equal to the initial value of i.
Proof by induction:

Base case: If the first iteration doesn’t terminate, A[i] will be equal to some value that is less than or equal to x, and A[j] will be equal to x, so the loop invariant holds for the second iteration of the outer loop.

Inductive case: For the nth iteration, where n > 1, by induction, we know that at the beginning, i < j, and A[i] ≤ x, A[j] ≥ x. Let m, n be the values of i, j at the beginning of this iteration, we know at line 11:
 i ≤ n,
 j ≥ m,
 A[i] ≥ x,
 A[j] ≤ x.
If i < j at line 11, after swapping A[i] and A[j]. we have A[i] ≤ x, and A[j] ≥ x. Now the (n + 1)th iteration begins, the loop invariant still holds.
If i ≥ j at line 11, the algorithm terminates, so the loop invariant still holds.
c.
Since A[p‥r] contains at least two elements, we know that p < r.
First, we prove that the returned j is less than r:
At the beginning of the first iteration, i = p  1, and j = r + 1. Notice that line 6 and line 9 will be executed at least once, also, we have A[p] = x, so we know i must equal to p in line 11. If line 6 is executed only once, j will equal to p at line 11. Since p < r, we know i < j, so the second iteration of the outer loop will be executed, which will cause j being decremented to a value that is less than r. If line 6 is executed more than once, j is still decremented to a value that is less than r.
Then, we prove that the returned j is greater than or equal to i:
If the loop terminates at the first iteration, i will be equal to p, otherwise the second iteration will be executed, which will cause i being incremented to a value that is more than p.
d.
The outer loop has the following loop invariant:
 At the beginning of each iteration, elements in A[p‥i] is less than or equal to x, and elements in A[j‥r] is greater than or equal to x.
Proof by induction:

Base case: In the beginning of the first iteration, both A[p‥i] and A[j‥r] is empty, so the claim holds.

Inductive case: In each iteration of the outer loop, Let m, n be the initial values of i, j. In the loop body, we find the fist element from the right that is less than or equal to x, and the first element from the left that is greater than or equal to x, then swap them. After swapping, we know that every element in A[m + 1‥i] is less than or equal to x, and every element in A[j + 1‥n] is greater than or equal to x. By induction, we know that A[p‥m] is less than or equal to x, and A[n‥r] is greater than or equal to x. So elements in A[p‥i] is less than or equal to x, and elements in A[j‥r] is greater than or equal to x at the beginning of next iteration.
If the algorithm terminates, at line 11, we have i ≥ j, and elements in A[p‥i  1] is less than or equal to x, and elements in A[j + 1‥r] is greater than or equal to x. So there are two cases:
┌─────┬─────┬─────┬─────┬─────┐ │ p │ … │ i j │ … │ r │ ├─────┼─────┼─────┼─────┼─────┤ │ ≤ x │ ≤ x │ = x │ ≥ x │ ≥ x │ └─────┴─────┴─────┴─────┴─────┘ ┌─────┬─────┬─────┬─────┬─────┬─────┐ │ p │ … │ j │ i │ … │ r │ ├─────┼─────┼─────┼─────┼─────┼─────┤ │ ≤ x │ ≤ x │ ≤ x │ ≥ x │ ≥ x │ ≥ x │ └─────┴─────┴─────┴─────┴─────┴─────┘
In both cases, elements in A[p‥j] is less than or equal to x, and elements in A[j + 1‥r] is greater than or equal to x.
e.
Solution is implemented here.
72 Quicksort with equal element values
The analysis of the expected running time of randomized quicksort in Section 7.4.2 assumes that all element values are distinct. In this problem, we examine what happens when they are not.
a. Suppose that all element values are equal. What would be randomized quicksort’s running time in this case?
b. The Partition procedure returns an index q such that each element of A[p‥q  1] is less than or equal to A[q] and each element of A[q + 1‥r] is greater than A[q]. Modify the Partition procedure to produce a procedure Partition′(A, p, r), which permutes the elements of A[p‥r] and returns two indices q and t, where p ≤ q ≤ t ≤ r, such that
 all elements of A[q‥t] are equal,
 each element of A[p‥q  1] is less than A[q], and
 each element of A[t + 1‥r] is greater than A[q].
Like Partition, your Partition′ procedure should take Θ(r  p) time.
c. Modify the RandomizedPartition procedure to call Partition′, and name the new procedure RandomizedQuicksort′. Then modify the Quicksort procedure to produce a procedure Quicksort′(A, p, r) that calls RandomizedPartition′ and recurses only on partitions of elements not known to be equal to each other.
d. Using Quicksort′, how would you adjust the analysis in Section 7.4.2 to avoid the assumption that all elements are distinct?
a.
Θ(\(n^2\)).
b.
Solution is implemented here.
c.
Solution is implemented here.
d.
Skipped.
73 Alternative quicksort analysis
An alternative analysis of the running time of randomized quicksort focuses on the expected running time of each individual recursive call to RandomizedQuicksort, rather than on the number of comparisons performed.
a. Argue that, given an array of size n, the probability that any particular element is chosen as the pivot is 1 / n. Use this to define indicator random variables \(X_i\) = I{ith smallest element is chosen as the pivot}. What is E[\(X_i\)]?
b. Let T(n) be a random variable denoting the running time of quicksort on an array of size n. Argue that
 E[T(n)] = E\(\displaystyle\left[∑_{q = 1}^n X_q (T(q  1) + T(n  q) + Θ(n))\right]\). (7.5)
c. Show that we can rewrite equation (7.5) as
 E[T(n)] = \(\displaystyle\frac{2}{n} ∑_{q = 2}^{n  1} \operatorname{E}\left[T(q)\right] + Θ(n)\). (7.6)
d. Show that
 \(\displaystyle ∑_{k = 2}^{n  1} k \lg k ≤ \frac{1}{2} n^2 \lg n  \frac{1}{8} n^2\). (7.7)
(Hint: Split the summation into two parts, one for k = 2, 3, …, ⌈n / 2⌉  1 and one for k = ⌈n / 2⌉, …, n  1.)
e. Using the bound from equation (7.7), show that the recurrence in equation (7.6) has the solution E[T(n)] = Θ(n lg n). (Hint: Show, by substitution, that E[T(n)] ≤ a n lg n for sufficiently large n and for some positive constant a.)
Skipped.
74 Stack depth for quicksort
The Quicksort algorithm of Section 7.1 contains two recursive calls to itself. After Quicksort calls Partition, it recursively sorts the left subarray and then it recursively sorts the right subarray. The second recursive call in Quicksort is not really necessary; we can avoid it by using an iterative control structure. This technique, called tail recursion, is provided automatically by good compilers. Consider the following version of quicksort, which simulates tail recursion:
TailRecursiveQuicksort(A, p, r)
 while p < r
 // Partition and sort left subarray.
 q = Partition(A, p, r)
 TailRecursiveQuicksort(A, p, q  1)
 p = q + 1
a. Argue that TailRecursiveQuicksort(A, 1, A.length) correctly sorts the array A.
Compilers usually execute recursive procedures by using a stack that contains pertinent information, including the parameter values, for each recursive call. The information for the most recent call is at the top of the stack, and the information for the initial call is at the bottom. Upon calling a procedure, its information is pushed onto the stack; when it terminates, its information is popped. Since we assume that array parameters are represented by pointers, the information for each procedure call on the stack requires O(1) stack space. The stack depth is the maximum amount of stack space used at any time during a computation.
b. Describe a scenario in which TailRecursiveQuicksort’s stack depth is Θ(n) on an nelement input array.
c. Modify the code for TailRecursiveQuicksort so that the worstcase stack depth is Θ(lg n). Maintain the O(n lg n) expected running time of the algorithm.
a.
The loop has the following invariant:
 Let m be the initial value of p. At the beginning of each iteration of the loop, A[m‥p  1] is in sorted order, and no elements in A[m‥p  1] is more than A[p].
Proof by induction:

Base case: At the beginning of the first iteration, p = m, so A[m‥p  1] is empty, the invariant holds.

Inductive case: At the beginning of the nth iteration, where n ≥ 1, by induction, we know that A[m‥p  1] is in sorted order, and no elements in A[m‥p  1] is more than A[p].
After calling Partition in line 3, we know that elements in A[p‥q  1] is less than or equal to A[q], and elements in A[q + 1‥r] is greater than A[q].
After calling TailRecursiveQuicksort in line 4, we know that A[p‥q  1] are in sorted order. Since no elements in A[m‥p  1] is more than A[p], A[m‥q] is in sorted order. After setting p = q + 1 in line 5, the next iteration begins, so the invariant still holds.
When the loop terminates, according to the loop invariant, we know that subarray A[p‥r] will be in sorted order.
b.
When the array is in increasing order.
c.
Solution is implemented here.
75 Medianof3 partition
One way to improve the RandomizedQuicksort procedure is to partition around a pivot that is chosen more carefully than by picking a random element from the subarray. One common approach is the medianof3 method: choose the pivot as the median (middle element) of a set of 3 elements randomly selected from the subarray. (See Exercise 7.46.) For this problem, let us assume that the elements in the input array A[1‥n] are distinct and that n ≥ 3. We denote the sorted output array by A′[1‥n]. Using the medianof3 method to choose the pivot element x, define \(p_i\) = Pr{x = A′[i]}.
a. Give an exact formula for \(p_i\) as a function of n and i for i = 2, 3, …, n  1. (Note that \(p_1 = p_n = 0\).)
b. By what amount have we increased the likelihood of choosing the pivot as x = A′[⌊(n + 1) / 2⌋], the median of A[1‥n], compared with the ordinary implementation? Assume that n → ∞, and give the limiting ratio of these probabilities.
c. If we define a “good” split to mean choosing the pivot as x = A′[i], where n / 3 ≤ i ≤ 2 n / 3, by what amount have we increased the likelihood of getting a good split compared with the ordinary implementation? (Hint: Approximate the sum by an integral.)
d. Argue that in the Ω(n lg n) running time of quicksort, the medianof3 method affects only the constant factor.
Skipped.
The medianof3 quicksort is implemented here.
76 Fuzzy sorting of intervals
Consider a sorting problem in which we do not know the numbers exactly. Instead, for each number, we know an interval on the real line to which it belongs. That is, we are given n closed intervals of the form [\(a_i\), \(b_i\)], where \(a_i\) ≤ \(b_i\). We wish to fuzzysort these intervals, i.e., to produce a permutation ⟨\(i_1\), \(i_2\), …, \(i_n\)⟩ of the intervals such that for j = 1, 2, …, n, there exist \(c_j\) ∈ [\(a_{i_j}\), \(a_{i_j}\)] satisfying \(c_1\) ≤ \(c_2\) ≤ ⋯ ≤ \(c_n\).
a. Design a randomized algorithm for fuzzysorting n intervals. Your algorithm should have the general structure of an algorithm that quicksorts the left endpoints (the \(a_i\) values), but it should take advantage of overlapping intervals to improve the running time. (As the intervals overlap more and more, the problem of fuzzysorting the intervals becomes progressively easier. Your algorithm should take advantage of such overlapping, to the extent that it exists.)
b. Argue that your algorithm runs in expected time Θ(n lg n) in general, but runs in expected time Θ(n) when all of the intervals overlap (i.e., when there exists a value x such that x ∈ [\(a_i\), \(b_i\)] for all i). Your algorithm should not be checking for this case explicitly; rather, its performance should naturally improve as the amount of overlap increases.
a. Solution is implemented here.
b. Skipped.
8 Sorting in Linear Time
8.1 Lower bounds for sorting
8.11
What is the smallest possible depth of a leaf in a decision tree for a comparison sort?
n  1.
Because for any element in the array to be sorted, it should be compared at lease once, otherwise we couldn’t determine its position in the sorted array. So if we must compare every element at least once, we need at least n  1 comparisons.
8.12
Obtain asymptotically tight bounds on lg(n!) without using Stirling’s approximation. Instead, evaluate the summation \(∑_{k = 1}^{n} \lg k\) using techniques from Section A.2.
Skipped.
8.13
Show that there is no comparison sort whose running time is linear for at least half of the n! inputs of length n. What about a fraction of 1 / n of the inputs of length n? What about a fraction 1 / \(2^n\)?
Skipped.
8.14
Suppose that you are given a sequence of n elements to sort. The input sequence consists of n / k subsequences, each containing k elements. The elements in a given subsequence are all smaller than the elements in the succeeding subsequence and larger than the elements in the preceding subsequence. Thus, all that is needed to sort the whole sequence of length n is to sort the k elements in each of the n / k subsequences. Show an Ω(n lg k) lower bound on the number of comparisons needed to solve this variant of the sorting problem. (Hint: It is not rigorous to simply combine the lower bounds for the individual subsequences.)
Skipped.
8.2 Counting sort
8.21
Using Figure 8.2 as a model, illustrate the operation of CountingSort on the array A = ⟨6, 0, 2, 0, 1, 3, 4, 6, 1, 3, 2⟩.
Skipped.
8.22
Prove that CountingSort is stable.
The key is the reverse iterating of line 10.
For all elements in A that equals to k, the last one will be put at the position C[k], and then C[k] will be decremented by one, so the second last element will be put before the last element and so on. So the relative order of equal elements is kept, which means the algorithm is stable.
8.23
Suppose that we were to rewrite the for loop header in line 10 of the CountingSort as
10 for j = 1 to A.length
Show that the algorithm still works properly. Is the modified algorithm stable?
It is the same as the original algorithm except that the order of the elements with the same value is reversed. The modified algorithm is not stable.
The algorithm is implemented here.
8.24
Describe an algorithm that, given n integers in the range 0 to k, preprocesses its input and then answers any query about how many of the n integers fall into a range [a‥b] in O(1) time. Your algorithm should use Θ(n + k) preprocessing time.
Solution is implemented here.
8.3 Radix sort
8.31
Using Figure 8.3 as a model, illustrate the operation of RadixSort on the following list of English words: COW, DOG, SEA, RUG, ROW, MOB, BOX, TAB, BAR, EAR, TAR, DIG, BIG, TEA, NOW, FOX.
Skipped.
8.32
Which of the following sorting algorithms are stable: insertion sort, merge sort, heapsort, and quicksort? Give a simple scheme that makes any comparison sort stable. How much additional time and space does your scheme entail?

Insertion sort and merge sort are stable.

First, convert each element A[i] to the tuple (A[i], i), then sort the converted array. When comparing two tuples, first compare the first element. If they equal to each other, compare the second element.
Both additional time and additional space are Θ(n).
The algorithm is implemented here.
8.33
Use induction to prove that radix sort works. Where does your proof need the assumption that the intermediate sort is stable?
The loop invariant:
 After the ith iteration, array A is sorted on the i lowestorder digits.
Proof by induction:
 Base case: After the first iteration, array A is being sorted on the first digit of its elements, so the array is sorted on the lowestorder digit.
 Inductive case: At the (i + 1)th iteration, the array is being sorted on the (i + 1)th lowestorder digit with a stable sort, which means the elements with the same (i + 1)th lowestorder digit will be grouped together in a sorted order; also, because the sort is stable, the relative order of the elements in any group is kept. By induction, we know that after the ith iteration, array A is sorted on the i lowestorder digits, so the elements in each group is sorted, which means the whole array is sorted on the (i + 1) lowestorder digits.
8.34
Show how to sort n integers in the range 0 to \(n^3\)  1 in O(n) time.
Represent the integers with lg n bit digits, then the running time will be:
Θ(d / r (n + \(2^r\)))
=Θ(lg \(n^3\) / lg n (n + \(2^{\lg n}\)))
= Θ(6 n)
= O(n).
8.35 ★
In the first cardsorting algorithm in this section, exactly how many sorting passes are needed to sort ddigit decimal numbers in the worst case? How many piles of cards would an operator need to keep track of in the worst case?
 \(∑_{i = 0}^{d  1} 10^i\) = (\(10^d  1\)) / 9.
 9 d.
8.4 Bucket sort
8.41
Using Figure 8.4 as a model, illustrate the operation of BucketSort on the array A = ⟨.79, .13, .16, .64, .39, .20, .89, .53, .71, .42⟩.
Skipped.
8.42
Explain why the worstcase running time for bucket sort is Θ(\(n^2\)). What simple change to the algorithm preserves its linear averagecase running time and makes its worstcase running time O(n lg n)?
 If all elements are put into one single bucket, the running time become Θ(\(n^2\)).
 Use merge sort instead of insertion sort to sort elements in buckets.
8.43
Let X be a random variable that is equal to the number of heads in two flips of a fair coin. What is E[\(X^2\)]? What is \(\operatorname{E}^2[X]\)?
E[\(X^2\)]
= \(0^2\) × 1 / 4 + \(1^2\) × 1 / 2 + \(2^2\) × 1 / 4
= 3 / 2.
\(\operatorname{E}^2[X]\)
= \((0 × 1 / 4 + 1 × 1 / 2 + 2 × 1 / 4)^2\)
= 1.
8.44 ★
We are given n points in the unit circle, \(p_i\) = (\(x_i\), \(y_i\)), such that 0 < \(x_i^2 + y_i^2\) ≤ 1 for i = 1, 2, …, n. Suppose that the points are uniformly distributed; that is, the probability of finding a point in any region of the circle is proportional to the area of that region. Design an algorithm with an averagecase running time of Θ(n) to sort the n points by their distances \(d_i = \sqrt{x_i^2 + y_i^2}\) from the origin. (Hint: Design the bucket sizes in BucketSort to reflect the uniform distribution of the points in the unit circle.)
Question: why can’t \(x_i^2 + y_i^2\) be zero?
Since the probability is proportional to the area, and the whole area is π, the we construct circles with area π / n, 2 π / n, …, (n  1) π / n, π, and use their boundaries as the bucket boundaries, that is we construct buckets with boundaries of \(\sqrt{1 / n}\), \(\sqrt{2 / n}\), …, \(\sqrt{(n  1) / n}\), 1.
To implement the algorithm, we need to put point (x, y), in bucket ⌈n (\(x^2 + y^2\))⌉  1.
The solution is implemented here.
8.45 ★
A probability distribution function P(x) for a random variable X is defined by P(x) = Pr{X ≤ x}. Suppose that we draw a list of n random variables \(X_1\), \(X_2\), …, \(X_n\) from a continuous probability distribution function P that is computable in O(1) time. Give an algorithm that sorts these numbers in linear averagecase time.
Using bucket sort by putting X into the bucket ⌈n P(X)⌉  1.
Proof: the probability of X going into bucket i is:
Pr{⌈n P(X)⌉  1 = i}
= Pr{⌈n P(X)⌉ = i + 1}
= Pr{i < n P(X) ≤ i + 1}
= Pr{i / n < P(X) ≤ (i + 1) / n}
= Pr{P(X) ≤ (i + 1) / n}  Pr{P(X) ≤ i / n}.
Since
Pr{P(X) ≤ t}
= Pr{X ≤ \(P^{1}(t)\)}
= \(P\left(P^{1}(t)\right)\)
= t,
Pr{P(X) ≤ (i + 1) / n}  Pr{P(X) ≤ i / n}
= (i + 1) / n  i / n
= 1 / n.
So the probability of X going into bucket i is 1 / n.
The solution is implemented here.
8.X Problems
81 Probabilistic lower bounds on comparison sorting
In this problem, we prove a probabilistic Ω(n lg n) lower bound on the running time of any deterministic or randomized comparison sort on n distinct input elements. We begin by examining a deterministic comparison sort A with decision tree \(T_A\). We assume that every permutation of A’s inputs is equally likely.
a. Suppose that each leaf of \(T_A\) is labeled with the probability that it is reached given a random input. Prove that exactly n! leaves are labeled 1 / n! and that the rest are labeled 0.
b. Let D(T) denote the external path length of a decision tree T; that is, D(T) is the sum of the depths of all the leaves of T. Let T be a decision tree with k > 1 leaves, and let LT and RT be the left and right subtrees of T. Show that D(T) = D(LT) + D(RT) + k.
c. Let d(k) be the minimum value of D(T) over all decision trees T with k > 1 leaves. Show that d(k) = \(\min_{1 ≤ i ≤ k  1} \left\lbrace d(i) + d(k  i) + k\right\rbrace\). (Hint: Consider a decision tree T with k leaves that achieves the minimum. Let \(i_0\) be the number of leaves in LT and k  \(i_0\) the number of leaves in RT.)
d. Prove that for a given value of k > 1 and i in the range 1 ≤ i ≤ k  1, the function i lg i + (k  i) lg(k  i) is minimized at i = k / 2. Conclude that d(k) = Ω(k lg k).
e. Prove that D(\(T_A\)) = Ω(n! lg(n!)), and conclude that the averagecase time to sort n elements is Ω(n lg n).
Now, consider a randomized comparison sort B. We can extend the decisiontree model to handle randomization by incorporating two kinds of nodes: ordinary comparison nodes and “randomization” nodes. A randomization node models a random choice of the form Random(1, r) made by algorithm B; the node has r children, each of which is equally likely to be chosen during an execution of the algorithm.
f. Show that for any randomized comparison sort B, there exists a deterministic comparison sort A whose expected number of comparisons is no more than those made by B.
Skipped.
82 Sorting in place in linear time
Suppose that we have an array of n data records to sort and that the key of each record has the value 0 or 1. An algorithm for sorting such a set of records might possess some subset of the following three desirable characteristics:
 The algorithm runs in O(n) time.
 The algorithm is stable.
 The algorithm sorts in place, using no more than a constant amount of storage space in addition to the original array.
a. Give an algorithm that satisfies criteria 1 and 2 above.
b. Give an algorithm that satisfies criteria 1 and 3 above.
c. Give an algorithm that satisfies criteria 2 and 3 above.
d. Can you use any of your sorting algorithms from parts (a)–(c) as the sorting method used in line 2 of RadixSort, so that RadixSort sorts n records with bbit keys in O(b n) time? Explain how or why not.
e. Suppose that the n records have keys in the range from 1 to k. Show how to modify counting sort so that it sorts the records in place in O(n + k) time. You may use O(k) storage outside the input array. Is your algorithm stable? (Hint: How would you do it for k = 3?)
a. Counting sort.
b. The Partition procedure used in quicksort, and use 0 as pivot.
c. Insertion sort.
d.
 (a) Counting sort: No. Because it requires Ω(\(2^b\)) running time.
 (b) Partition: No. It does not sort keys with more than one bit.
 (c) Insertion sort: Yes.
e.
The algorithm is implemented here. It is unstable.
83 Sorting variablelength items
a. You are given an array of integers, where different integers may have different numbers of digits, but the total number of digits over all the integers in the array is n. Show how to sort the array in O(n) time.
b. You are given an array of strings, where different strings may have different numbers of characters, but the total number of characters over all the strings is n. Show how to sort the strings in O(n) time.
(Note that the desired order here is the standard alphabetical order; for example,
a
<ab
<b
.)
a. We only consider positive integers without leading zeroes. For each integer k, we put it into the bucket i if it has i digits. Then radix sort each bucket. Finally concatenate elements in each bucket.
b. Solution is implemented here.
84 Water jugs
Suppose that you are given n red and n blue water jugs, all of different shapes and sizes. All red jugs hold different amounts of water, as do the blue ones. Moreover, for every red jug, there is a blue jug that holds the same amount of water, and vice versa.
Your task is to find a grouping of the jugs into pairs of red and blue jugs that hold the same amount of water. To do so, you may perform the following operation: pick a pair of jugs in which one is red and one is blue, fill the red jug with water, and then pour the water into the blue jug. This operation will tell you whether the red or the blue jug can hold more water, or that they have the same volume. Assume that such a comparison takes one time unit. Your goal is to find an algorithm that makes a minimum number of comparisons to determine the grouping. Remember that you may not directly compare two red jugs or two blue jugs.
a. Describe a deterministic algorithm that uses Θ(\(n^2\)) comparisons to group the jugs into pairs.
b. Prove a lower bound of Ω(n lg n) for the number of comparisons that an algorithm solving this problem must make.
c. Give a randomized algorithm whose expected number of comparisons is O(n lg n), and prove that this bound is correct. What is the worstcase number of comparisons for your algorithm?
a. For each red jug, we compare it with every blue jug, and find the one with the same volume.
b. It is like sorting, if our algorithm correctly groups jugs into pairs, there must be n! different grouping possible result. So the decision tree must at least have the height lg n! = Ω(n lg n).
c. Solution is implemented here. The worstcast running time is Θ(\(n^2\)).
85 Average sorting
Suppose that, instead of sorting an array, we just require that the elements increase on average. More precisely, we call an nelement array A ksorted if, for all i = 1, 2, …, n  k, the following holds:
\(\frac{∑_{j = i}^{i + k  1} A[j]}{k} ≤ \frac{∑_{j = i + 1}^{i + k} A[j]}{k}\).
a. What does it mean for an array to be 1sorted?
b. Give a permutation of the numbers 1, 2, …, 10 that is 2sorted, but not sorted.
c. Prove that an nelement array is ksorted if and only if A[i] ≤ A[i + k] for all i = 1, 2, …, n  k.
d. Give an algorithm that ksorts an nelement array in O(n lg(n / k)) time.
We can also show a lower bound on the time to produce a ksorted array, when k is a constant.
e. Show that we can sort a ksorted array of length n in O(n lg k) time. (Hint: Use the solution to Exercise 6.59.)
f. Show that when k is a constant, ksorting an nelement array requires Ω(n lg n) time. (Hint: Use the solution to the previous part along with the lower bound on comparison sorts.)
a.
It means the array is sorted.
b.
⟨2, 1, 4, 3, 6, 5, 8, 7, 10, 9⟩.
c.
\(\frac{∑_{j = i}^{i + k  1} A[j]}{k} ≤ \frac{∑_{j = i + 1}^{i + k} A[j]}{k}\)
⇔ \(∑_{j = i}^{i + k  1} A[j] ≤ ∑_{j = i + 1}^{i + k} A[j]\)
⇔ \(A[i] + ∑_{j = i + 1}^{i + k  1} A[j] ≤ ∑_{j = i + 1}^{i + k  1} A[j] + A\left[i + k\right]\)
⇔ A[i] ≤ A[i + k].
d.
Running an O(n lg n) running time sorting algorithm on subarrays ⟨A[j], A[k + j], A[2 k + j], …⟩ for j = 0, 1, …, k  1.
e.
A ksorted array contains at most k sorted subarrays of length n / k, we can merge k sorted arrays in O(n / k lg k) = O(n lg k) time.
f.
Proof by contradiction: If ksorting an nelement array requires o(n lg n) time, then sorting an n element will only need o(n lg n) + O(n lg k) = o(n lg n) time, which is impossible. So ksorting an nelement array requires Ω(n lg n) time.
86 Lower bound on merging sorted lists
The problem of merging two sorted lists arises frequently. We have seen a procedure for it as the subroutine Merge in Section 2.3.1. In this problem, we will prove a lower bound of 2 n  1 on the worstcase number of comparisons required to merge two sorted lists, each containing n items.
First we will show a lower bound of 2 n  o(n) comparisons by using a decision tree.
a. Given 2 n numbers, compute the number of possible ways to divide them into two sorted lists, each with n numbers.
b. Using a decision tree and your answer to part (a), show that any algorithm that correctly merges two sorted lists must perform at least 2 n  o(n) comparisons.
Now we will show a slightly tighter 2 n  1 bound.
c. Show that if two elements are consecutive in the sorted order and from different lists, then they must be compared.
d. Use your answer to the previous part to show a lower bound of 2 n  1 comparisons for merging two sorted lists.
a.
C(2 n, n) = \(\frac{(2 n)!}{(n!)^2}\).
b.
The decision tree has at least \(\frac{(2 n)!}{(n!)^2}\) nodes, so the height of the tree is at least
\(\lg \frac{(2 n)!}{(n!)^2}\)
= lg ((2 n)!)  2 lg(n!)
= \(∑_{i = 1}^{2 n} \lg i  2 ∑_{i = 1}^{n} \lg i\)
= \(∑_{i = n + 1}^{2 n} \lg i  ∑_{i = 1}^{n} \lg i\)
= \(∑_{i = 1}^{n} \lg (n + i)  ∑_{i = 1}^{n} \lg i\)
= \(∑_{i = 1}^{n} \lg (1 + n / i)\).
Skipped.
c.
Skipped.
d.
Skipped.
87 The 01 sorting lemma and columnsort
A compareexchange operation on two array elements A[i] and A[j], where i < j, has the form
CompareExchange(A, i, j)
 if A[i] > A[j]
 exchange A[i] with A[j]
After the compareexchange operation, we know that A[i] ≤ A[j].
An oblivious compareexchange algorithm operates solely by a sequence of prespecified compareexchange operations. The indices of the positions compared in the sequence must be determined in advance, and although they can depend on the number of elements being sorted, they cannot depend on the values being sorted, nor can they depend on the result of any prior compareexchange operation. For example, here is insertion sort expressed as an oblivious compareexchange algorithm:
InsertionSort(A)
 for j = 2 to A.length
 for i = j  1 downto 1
 CompareExchange(A, i, i + 1)
The 01 sorting lemma provides a powerful way to prove that an oblivious compareexchange algorithm produces a sorted result. It states that if an oblivious compareexchange algorithm correctly sorts all input sequences consisting of only 0s and 1s, then it correctly sorts all inputs containing arbitrary values.
You will prove the 01 sorting lemma by proving its contrapositive: if an oblivious compareexchange algorithm fails to sort an input containing arbitrary values, then it fails to sort some 01 input. Assume that an oblivious compareexchange algorithm X fails to correctly sort the array A[1‥n]. Let A[p] be the smallest value in A that algorithm X puts into the wrong location, and let A[q] be the value that algorithm X moves to the location into which A[p] should have gone. Define an array B[1‥n] of 0s and 1s as follows:
B[i] = \(\begin{cases} 0 &&\text{if $A[i] ≤ A[p]$,}\\ 1 &&\text{if $A[i] > A[p]$.} \end{cases}\)
a. Argue that A[q] > A[p], so that B[p] = 0 and B[q] = 1.
b. To complete the proof of the 01 sorting lemma, prove that algorithm X fails to sort array B correctly.
Now you will use the 01 sorting lemma to prove that a particular sorting algorithm works correctly. The algorithm, columnsort, works on a rectangular array of n elements. The array has r rows and s columns (so that n = r s), subject to three restrictions:
 r must be even,
 s must be a divisor of r, and
 r ≥ 2 \(s^2\).
When columnsort completes, the array is sorted in columnmajor order: reading down the columns, from left to right, the elements monotonically increase.
Columnsort operates in eight steps, regardless of the value of n. The odd steps are all the same: sort each column individually. Each even step is a fixed permutation. Here are the steps:
 Sort each column.
 Transpose the array, but reshape it back to r rows and s columns. In other words, turn the leftmost column into the top r / s rows, in order; turn the next column into the next r / s rows, in order; and so on.
 Sort each column.
 Perform the inverse of the permutation performed in step 2.
 Sort each column.
 Shift the top half of each column into the bottom half of the same column, and shift the bottom half of each column into the top half of the next column to the right. Leave the top half of the leftmost column empty. Shift the bottom half of the last column into the top half of a new rightmost column, and leave the bottom half of this new column empty.
 Sort each column.
 Perform the inverse of the permutation performed in step 6.
Figure 8.5 shows an example of the steps of columnsort with r = 6 and s = 3. (Even though this example violates the requirement that r ≥ 2 \(s^2\), it happens to work.)
10 14 5 4 1 2 4 8 10 1 3 6 1 4 11 8 7 17 8 3 5 12 16 18 2 5 7 3 8 14 12 1 6 10 7 6 1 3 7 4 8 10 6 10 17 16 9 11 12 9 11 9 14 15 9 13 15 2 9 12 4 15 2 16 14 13 2 5 6 11 14 17 5 13 16 18 3 13 18 15 17 11 13 17 12 16 18 7 15 18 (a) (b) (c) (d) (e) 1 4 11 5 10 16 4 10 16 1 7 13 2 8 12 6 13 17 5 11 17 2 8 14 3 9 14 7 15 18 6 12 18 3 9 15 5 10 16 1 4 11 1 7 13 4 10 16 6 13 17 2 8 12 2 8 14 5 11 17 7 15 18 3 9 14 3 9 15 6 12 18 (f) (g) (h) (i)
Figure 8.5 The steps of columnsort. (a) The input array with 6 rows and 3 columns. (b) After sorting each column in step 1. (c) After transposing and reshaping in step 2. (d) After sorting each column in step 3. (e) After performing step 4, which inverts the permutation from step 2. (f) After sorting each column in step 5. (g) After shifting by half a column in step 6. (h) After sorting each column in step 7. (i) After performing step 8, which inverts the permutation from step 6. The array is now sorted in columnmajor order.
c. Argue that we can treat columnsort as an oblivious compareexchange algorithm, even if we do not know what sorting method the odd steps use.
Although it might seem hard to believe that columnsort actually sorts, you will use the 01 sorting lemma to prove that it does. The 01 sorting lemma applies because we can treat columnsort as an oblivious compareexchange algorithm. A couple of definitions will help you apply the 01 sorting lemma. We say that an area of an array is clean if we know that it contains either all 0s or all 1s. Otherwise, the area might contain mixed 0s and 1s, and it is dirty. From here on, assume that the input array contains only 0s and 1s, and that we can treat it as an array with r rows and s columns.
d. Prove that after steps 1–3, the array consists of some clean rows of 0s at the top, some clean rows of 1s at the bottom, and at most s dirty rows between them.
e. Prove that after step 4, the array, read in columnmajor order, starts with a clean area of 0s, ends with a clean area of 1s, and has a dirty area of at most \(s^2\) elements in the middle.
f. Prove that steps 5–8 produce a fully sorted 01 output. Conclude that columnsort correctly sorts all inputs containing arbitrary values.
g. Now suppose that s does not divide r. Prove that after steps 1–3, the array consists of some clean rows of 0s at the top, some clean rows of 1s at the bottom, and at most 2 s  1 dirty rows between them. How large must r be, compared with s, for columnsort to correctly sort when s does not divide r?
h. Suggest a simple change to step 1 that allows us to maintain the requirement that r ≥ 2 \(s^2\) even when s does not divide r, and prove that with your change, columnsort correctly sorts.
Skipped.
9 Medians and Order Statistics
9.1 Minimum and maximum
9.11
Show that the second smallest of n elements can be found with n + ⌈lg n⌉  2 comparisons in the worst case. (Hint: Also find the smallest element.)
Solution is implemented here.
9.12 ★
Prove the lower bound of ⌈3 n / 2⌉  2 comparisons in the worst case to find both the maximum and minimum of n numbers. (Hint: Consider how many numbers are potentially either the maximum or minimum, and investigate how a comparison affects these counts.)
Skipped.
9.2 Selection in expected linear time
9.21
Show that RandomizedSelect never makes a recursive call to a 0length array.

At line 8, we know that:
i < k
⇒ i < q  p + 1
⇒ 1 < q  p + 1
⇒ (q  1)  p > 1
⇒ (q  1)  p ≥ 0.So the recursion call in line 8 is on a nonempty subarray A[p‥q  1].

At line 9, we know that:
i > k
⇒ i > q  p + 1
⇒ r  p + 1 > q  p + 1
⇒ r  (q + 1) > 1
⇒ r  (q + 1) ≥ 0.So the recursion call in line 8 is on a nonempty subarray A[q + 1‥r].
9.22
Argue that the indicator random variable \(X_k\) and the value T(max(k  1, n  k)) are independent.
Skipped.
9.23
Write an iterative version of RandomizedSelect.
Solution is implemented here.
9.24
Suppose we use RandomizedSelect to select the minimum element of the array A = ⟨3, 2, 9, 0, 7, 5, 4, 8, 6, 1⟩. Describe a sequence of partitions that results in a worstcase performance of RandomizedSelect.
On each partition, the maximum element is selected as the pivot.
9.3 Selection in worstcase linear time
9.31
In the algorithm Select, the input elements are divided into groups of 5. Will the algorithm work in linear time if they are divided into groups of 7? Argue that Select does not run in linear time if groups of 3 are used.
Assume the elements are divided into groups of k, we know that the number of elements greater than x is at least ((k + 1) / 2) (⌈⌈n / k⌉ / 2⌉  2) ≥ (k + 1) n / (4 k)  k  1. So in the worst cast, the recursive call in step 5 is on at most (3 k  1) n / (4 k) + k + 1. So for n that is large enough:
T(n)
≤ c ⌈n / k⌉ + c ((3 k  1) n / (4 k) + k + 1) + a n
≤ c n / k + c + c (3 k  1) n / (4 k) + c (k + 1) + a n
= (3 / 4) c (k + 1) n / k + c (k + 2) + a n
= (3 / 4) c (k + 1) n / k + c (k + 2) + a n
= c n + ((c ((3 / 4) (k + 1) / k  1) + a) n + c (k + 2)).
Solving c ((3 / 4) (k + 1) / k  1) + a < 0, we get:
c > a / (1  (3 / 4) (k + 1) / k).
 If k = 3, c > a / (1  (3 / 4) (3 + 1) / 3) = ∞, so that’s not possible.
 If k = 7, c > a / (1  (3 / 4) (7 + 1) / 7) = 7 a, so it is possible to select a c > 7 a to make T(n) linear.
9.32
Analyze Select to show that if n ≥ 140, then at least ⌈n / 4⌉ elements are greater than the medianofmedians x and at least ⌈n / 4⌉ elements are less than x.
Skipped.
9.33
Show how quicksort can be made to run in O(n lg n) time in the worst case, assuming that all elements are distinct.
When doing partition, use the Select procedure to find the median element as the pivot.
9.34 ★
Suppose that an algorithm uses only comparisons to find the ith smallest element in a set of n elements. Show that it can also find the i  1 smaller elements and the n  i larger elements without performing any additional comparisons.
Skipped.
9.35
Suppose that you have a “blackbox” worstcase lineartime median subroutine. Give a simple, lineartime algorithm that solves the selection problem for an arbitrary order statistic.
Replace the first 3 steps of Select to a step of using the lineartime median subroutine to find the pivot.
The running time becomes:
T(n) = T(n / 2) + O(n).
Using the master theorem, we have T(n) = O(n).
9.36
The kth quantiles of an nelement set are the k  1 order statistics that divide the sorted set into k equalsized sets (to within 1). Give an O(n lg k)time algorithm to list the kth quantiles of a set.
Skipped.
9.37
Describe an O(n)time algorithm that, given a set S of n distinct numbers and a positive integer k ≤ n, determines the k numbers in S that are closest to the median of S.
Skipped.
9.38
Let X[1‥n] and Y[1‥n] be two arrays, each containing n numbers already in sorted order. Give an O(lg n)time algorithm to find the median of all 2 n elements in arrays X and Y.
Solution is implemented here.
9.39
Professor Olay is consulting for an oil company, which is planning a large pipeline running east to west through an oil field of n wells. The company wants to connect a spur pipeline from each well directly to the main pipeline along a shortest route (either north or south), as shown in Figure 9.2. Given the x and ycoordinates of the wells, how should the professor pick the optimal location of the main pipeline, which would be the one that minimizes the total length of the spurs? Show how to determine the optimal location in linear time.
Well Well │ │ Well │ Well │ │ │ Well │ ──┬┴────┴───┬───┴────┴──┬───┴──┬── │ │ │ │ Well │ Well │ Well Well
Figure 9.2 Professor Olay needs to determine the position of the eastwest oil pipeline that minimizes the total length of the northsouth spurs.
Notice that the xcoordinates does not affect the total length of the northsouth spurs, so we only consider the ycoordinates.
First, the conclusion: we should put the pipeline anywhere between the lower median and upper median of the ycoordinates of the wells.
Proof:
If we sort the ycoordinates of the wells by into an array A[1‥n], let k be the y coordinates of the pipeline, then total length of the spurs is:
\(∑_{i = 1}^n A[i]  k\)
= \(∑_{i = 1}^{\lfloor n / 2\rfloor} A[i]  k + ∑_{i = 1}^{\lfloor n / 2\rfloor} A[n + 1  i]  k\) +
(n mod 2) A[⌈n / 2⌉]  k
= \(∑_{i = 1}^{\lfloor n / 2\rfloor} (A[i]  k + A[n + 1  i]  k)\) + (n mod 2) A[⌈n / 2⌉]  k.
If A[i] ≤ k ≤ A[n + 1  i], A[i]  k + A[n + 1  i]  k will have the minimal value of A[n + 1  i]  A[i], which means k should between A[i] and A[n + 1  i]. If A[⌊n / 2⌋] ≤ A[k] ≤ A[n + 1  ⌊n / 2⌋], k is between every (A[i], A[n + 1  i]) pair of wells. If n is even, we can put the pipeline anywhere between A[n / 2] and A[n / 2 + 1]; if n is odd, we must put the pipeline at coordinate A[(n + 1) / 2].
So we only need to use the Select procedure to find the median of all ycoordinates of the wells, which takes linear time.
9.X Problems
91 Largest i numbers in sorted order
Given a set of n numbers, we wish to find the i largest in sorted order using a comparisonbased algorithm. Find the algorithm that implements each of the following methods with the best asymptotic worstcase running time, and analyze the running times of the algorithms in terms of n and i.
a. Sort the numbers, and list the i largest.
b. Build a maxpriority queue from the numbers, and call ExtractMax i times.
c. Use an orderstatistic algorithm to find the ith largest number, partition around that number, and sort the i largest numbers.
a. Merge sort, then list the i largest. Running time is O(n lg n + i).
b. Running time is O(n + i lg n).
c. Use Select to find the ith largest number, then partition around the ith largest number, then merge sort i largest numbers. Running time is O(n + i lg i).
92 Weighted median
For n distinct elements \(x_1\), \(x_2\), …, \(x_n\) with positive weights \(w_1\), \(w_2\), …, \(w_n\) such that \(∑_{i = 1}^n w_i\) = 1, the weighted (lower) median is the element \(x_k\) satisfying
\(\displaystyle ∑_{x_i < x_k} w_i < \frac{1}{2}\)
and
\(\displaystyle ∑_{x_i > x_k} w_i ≤ \frac{1}{2}\).
For example, if the elements are 0.1, 0.35, 0.05, 0.1, 0.15, 0.05, 0.2 and each element equals its weight (that is, \(w_i = x_i\) for i = 1, 2, …, 7), then the median is 0.1, but the weighted median is 0.2.
a. Argue that the median of \(x_1\), \(x_2\), …, \(x_n\) is the weighted median of the \(x_i\) with weights \(w_i\) = 1 / n for i = 1, 2, …, n.
b. Show how to compute the weighted median of n elements in O(n lg n) worstcase time using sorting.
c. Show how to compute the weighted median in Θ(n) worstcase time using a lineartime median algorithm such as Select from Section 9.3.
The postoffice location problem is defined as follows. We are given n points \(p_1\), \(p_2\), …, \(p_n\) with associated weights \(w_1\), \(w_2\), …, \(w_n\). We wish to find a point p (not necessarily one of the input points) that minimizes the sum \(∑_{i = 1}^n w_i d\left(p, p_i\right)\), where d(a, b) is the distance between points a and b.
d. Argue that the weighted median is a best solution for the 1dimensional postoffice location problem, in which points are simply real numbers and the distance between points a and b is d(a, b) = a  b.
e. Find the best solution for the 2dimensional postoffice location problem, in which the points are (x, y) coordinate pairs and the distance between points a = (\(x_1\), \(y_1\)) and b = (\(x_2\), \(y_2\)) is the Manhattan distance given by d(a, b) = \(x_1  x_2\) + \(y_1  y_2\).
a.
4 2 5 2 6 3 7 3
There are ⌊(n  1) / 2⌋ elements that are less than the median, and ⌊n / 2⌋ elements that are greater than the median. So we have
 \(∑_{x_i < x_k} w_i\) = ⌊(n  1) / 2⌋ / n < 1 / 2,
 \(∑_{x_i > x_k} w_i\) = ⌊n / 2⌋ / n ≤ 1 / 2.
That satisfies the weighted median condition.
b.
First, sort the elements by their value, then accumulate the weight in the sorted array from left to right, until the accumulated value is greater than or equal to 1 / 2. Then the weighted median is the first element that makes the accumulated value greater than or equal to 1 / 2.
c.
The solution is implemented here.
d.
Skipped.
e.
\(∑_{i = 1}^n w_i d\left(p, p_i\right)\)
= \(∑_{i = 1}^n w_i \left(\leftx_1  x_2\right + \lefty_1  y_2\right\right)\)
= \(∑_{i = 1}^n w_i \leftx_1  x_2\right + ∑_{i = 1}^n w_i \lefty_1  y_2\right\)
So we can find the weight medians of xcoordinates and ycoordinates individually, then combine the result into a new coordinates, which should be the location of the postoffice.
93 Small order statistics
We showed that the worstcase number T(n) of comparisons used by Select to select the ith order statistic from n numbers satisfies T(n) = Θ(n), but the constant hidden by the Θnotation is rather large. When i is small relative to n, we can implement a different procedure that uses Select as a subroutine but makes fewer comparisons in the worst case.
a. Describe an algorithm that uses \(U_i(n)\) comparisons to find the ith smallest of n elements, where
\(U_i(n)\begin{cases} T(n)&\text{if $i ≥ n / 2$,}\\ \left\lfloor n / 2\right\rfloor + U_i\left(\left\lceil n / 2\right\rceil\right) + T(2 i)&\text{otherwise}. \end{cases}\)
(Hint: Begin with ⌊n / 2⌋ disjoint pairwise comparisons, and recurse on the set containing the smaller element from each pair.)
b. Show that, if i < n / 2, then \(U_i(n)\) = n + O(T(2 i / lg(n / i)).
c. Show that if i is a constant less than n / 2, then \(U_i(n)\) = n + O(lg n).
d. Show that if i = n / k for k ≥ 2, then \(U_i\)(n) = n + O(T(2 n / k) lg k).
Skipped.
94 Alternative analysis of randomized selection
In this problem, we use indicator random variables to analyze the RandomizedSelect procedure in a manner akin to our analysis of RandomizedQuicksort in Section 7.4.2.
As in the quicksort analysis, we assume that all elements are distinct, and we rename the elements of the input array A as \(z_1\), \(z_2\), …, \(z_n\), where \(z_i\) is the ith smallest element. Thus, the call RandomizedSelect(A, 1, n, k) returns \(z_k\).
For 1 ≤ i < j ≤ n, let
\(X_{i j k}\) = I{\(z_i\) is compared with \(z_j\) sometime during the execution of the algorithm to find \(z_k\)}.
a. Give an exact expression for E[\(X_{i j k}\)]. (Hint: Your expression may have different values, depending on the values of i, j, and k.)
b. Let \(X_k\) denote the total number of comparisons between elements of array A when finding \(z_k\). Show that
\(\displaystyle \operatorname{E}\left[X_k\right] ≤ 2 \left(∑_{i = 1}^k ∑_{j = k}^n \frac{1}{j  i + 1} + ∑_{j = k + 1}^n \frac{j  k  1}{j  k + 1} + ∑_{i = 1}^{k  2} \frac{k  i  1}{k  i + 1}\right)\).
c. Show that E[\(X_k\)] ≤ 4 n.
d. Conclude that, assuming all elements of array A are distinct, RandomizedSelect runs in expected time O(n).
Skipped.
III Data Structures
10 Elementary Data Structures
10.1 Stacks and queues
10.11
Using Figure 10.1 as a model, illustrate the result of each operation in the sequence Push(S, 4), Push(S, 1), Push(S, 3), Pop(S), Push(S, 8), and Pop(S) on an initially empty stack S stored in array S[1‥6].
Skipped.
10.12
Explain how to implement two stacks in one array A[1…n] in such a way that neither stack overflows unless the total number of elements in both stacks together is n. The Push and Pop operations should run in O(1) time.
Solution is implemented here.
10.13
Using Figure 10.2 as a model, illustrate the result of each operation in the sequence Enqueue(Q, 4), Enqueue(Q, 1), Enqueue(Q, 3), Dequeue(Q), Enqueue(Q, 8), and Dequeue(Q) on an initially empty queue Q stored in array Q[1‥6].
Skipped.
10.14
Rewrite Enqueue and Dequeue to detect underflow and overflow of a queue.
 Underflow condition: Q.tail == Q.head.
 Overflow condition: (Q.tail + 1) mod Q.length + 1 == Q.head.
10.15
Whereas a stack allows insertion and deletion of elements at only one end, and a queue allows insertion at one end and deletion at the other end, a deque (doubleended queue) allows insertion and deletion at both ends. Write four O(1)time procedures to insert elements into and delete elements from both ends of a deque implemented by an array.
here.
10.16
Show how to implement a queue using two stacks. Analyze the running time of the queue operations.
Solution is implemented here.
Push need O(1) running time, Pop need O(1) running time on average, but has worstcase O(n) running time.
10.17
Show how to implement a stack using two queues. Analyze the running time of the stack operations.
Solution is implemented here.
Push need O(1) running time, Pop need O(n) running time.
10.2 Linked lists
10.21
Can you implement the dynamicset operation Insert on a singly linked list in O(1) time? How about Delete?
Insert can be done in O(1) time.
As for Delete, if the node to be deleted is the last node, I will have to use O(n) time to delete it, but for nodes that are not the last node, I can move the contents of the next node into the node to be deleted, which will make the linked list look like the node is deleted. This method takes O(1) time.
Solution is implemented here.
10.22
Implement a stack using a singly linked list L. The operations Push and Pop should still take O(1) time.
Solution is implemented here.
10.23
Implement a queue by a singly linked list L. The operations Enqueue and Dequeue should still take O(1) time.
Solution is implemented here.
10.24
As written, each loop iteration in the ListSearch′ procedure requires two tests: one for x ≠ L.nil and one for x.key ≠ k. Show how to eliminate the test for x ≠ L.nil in each iteration.
Set the value of x.nil.key to k, so that if every test fails for k, we will get x.nil as the result, which is exactly what we wanted.
10.25
Implement the dictionary operations Insert, Delete, and Search using singly linked, circular lists. What are the running times of your procedures?
Skipped.
10.26
The dynamicset operation Union takes two disjoint sets \(S_1\) and \(S_2\) as input, and it returns a set S = \(S_1 ∪ S_2\) consisting of all the elements of \(S_1\) and \(S_2\). The sets \(S_1\) and \(S_2\) are usually destroyed by the operation. Show how to support Union in O(1) time using a suitable list data structure.
Both singly and doubly linked list are fine, just stitch the tail of the first list onto the head of the second list and you are done.
10.27
Give a Θ(n)time nonrecursive procedure that reverses a singly linked list of n elements. The procedure should use no more than constant storage beyond that needed for the list itself.
Remember that we have implemented a stack using singly linked list in exercise 10.22? To reverse a singly linked list, just treat the list as a stack, and pop their elements to construct a new list, the newly constructed list will be the reversion of the original list.
10.28 ★
Explain how to implement doubly linked lists using only one pointer value x.np per item instead of the usual two (next and prev). Assume that all pointer values can be interpreted as kbit integers, and define x.np to be x.np = x.next XOR x.prev, the kbit “exclusiveor” of x.next and x.prev. (The value nil is represented by 0.) Be sure to describe what information you need to access the head of the list. Show how to implement the Search, Insert, and Delete operations on such a list. Also show how to reverse such a list in O(1) time.
Skipped.
10.3 Implementing pointers and objects
10.31
Draw a picture of the sequence ⟨13, 4, 8, 19, 5, 11⟩ stored as a doubly linked list using the multiplearray representation. Do the same for the singlearray representation.
Skipped.
10.32
Write the procedures AllocateObject and FreeObject for a homogeneous collection of objects implemented by the singlearray representation.
Solution is implemented here.
10.33
Why don’t we need to set or reset the prev attributes of objects in the implementation of the AllocateObject and FreeObject procedures?
We treat the free list as a singly linked list, which only uses the next attribute, we don’t care for attributes other than prev.
10.34
It is often desirable to keep all elements of a doubly linked list compact in storage, using, for example, the first m index locations in the multiplearray representation. (This is the case in a paged, virtualmemory computing environment.) Explain how to implement the procedures AllocateObject and FreeObject so that the representation is compact. Assume that there are no pointers to elements of the linked list outside the list itself. (Hint: Use the array implementation of a stack.)
Skipped.
10.35
Let L be a doubly linked list of length n stored in arrays key, prev, and next of length m. Suppose that these arrays are managed by AllocateObject and FreeObject procedures that keep a doubly linked free list F. Suppose further that of the m items, exactly n are on list L and m  n are on the free list. Write a procedure CompactifyList(L, F) that, given the list L and the free list F, moves the items in L so that they occupy array positions 1, 2, …, n and adjusts the free list F so that it remains correct, occupying array positions n + 1; n + 2, …, m. The running time of your procedure should be Θ(n), and it should use only a constant amount of extra space. Argue that your procedure is correct.
Solution is implemented here.
10.4 Representing rooted trees
10.41
Draw the binary tree rooted at index 6 that is represented by the following attributes:
index key left right 1 12 7 3 2 15 8 nil 3 4 10 nil 4 10 5 9 5 2 nil nil 6 18 1 4 7 7 nil nil 8 14 6 2 9 21 nil nil 10 5 nil nil
Skipped.
10.42
Write an O(n)time recursive procedure that, given an nnode binary tree, prints out the key of each node in the tree.
Solution is implemented here.
10.43
Write an O(n)time nonrecursive procedure that, given an nnode binary tree, prints out the key of each node in the tree. Use a stack as an auxiliary data structure.
Solution is implemented here.
10.44
Write an O(n)time procedure that prints all the keys of an arbitrary rooted tree with n nodes, where the tree is stored using the leftchild, rightsibling representation.
Solution is implemented here.
Notice that the solution is exactly the same as exercise 10.42, because the structure of an unbounded branching tree with leftchild, rightsibling representation is the same as a binary tree, but the children of nodes has different meaning, so we can iterate through the tree with the same method used for iterating binary trees.
10.45 ★
Write an O(n)time nonrecursive procedure that, given an nnode binary tree, prints out the key of each node. Use no more than constant extra space outside of the tree itself and do not modify the tree, even temporarily, during the procedure.
Solution is implemented here.
10.46 ★
The leftchild, rightsibling representation of an arbitrary rooted tree uses three pointers in each node: leftchild, rightsibling, and parent. From any node, its parent can be reached and identified in constant time and all its children can be reached and identified in time linear in the number of children. Show how to use only two pointers and one boolean value in each node so that the parent of a node or all of its children can be reached and identified in time linear in the number of children.
Let the three members be leftchild, rightsiblingorparent and islastchild.
 If the node is the last child of its parent, store pointer to left child in leftchild, pointer to parent in rightsiblingorparent, and true in islastchild.
 Otherwise store pointer to left child in leftchild, pointer to right sibling in rightsiblingorparent, and true in islastchild.
10.X Problems
101 Comparisons among lists
For each of the four types of lists in the following table, what is the asymptotic worstcase running time for each dynamicset operation listed?
unsorted, singly linked sorted, singly linked unsorted, doubly linked sorted, doubly linked Search(L, k) Insert(L, x) Delete(L, x) Successor(L, x) Predecessor(L, x) Minimum(L) Maximum(L)
unsorted, singly linked  sorted, singly linked  unsorted, doubly linked  sorted, doubly linked  

Search(L, k)  O(n)  O(n)  O(n)  O(n) 
Insert(L, x)  O(1)  O(n)  O(1)  O(n) 
Delete(L, x)  O(n)  O(n)  O(1)  O(1) 
Successor(L, x)  O(1)  O(1)  O(1)  O(1) 
Predecessor(L, x)  O(n)  O(n)  O(1)  O(1) 
Minimum(L)  O(n)  O(1)  O(n)  O(1) 
Maximum(L)  O(n)  O(n)  O(n)  O(n) 
102 Mergeable heaps using linked lists
A mergeable heap supports the following operations: MakeHeap (which creates an empty mergeable heap), Insert, Minimum, ExtractMin, and Union. Show how to implement mergeable heaps using linked lists in each of the following cases. Try to make each operation as efficient as possible. Analyze the running time of each operation in terms of the size of the dynamic set(s) being operated on.
 a. Lists are sorted.
 b. Lists are unsorted.
 c. Lists are unsorted, and dynamic sets to be merged are disjoint.
Skipped.
103 Searching a sorted compact list
Exercise 10.34 asked how we might maintain an nelement list compactly in the first n positions of an array. We shall assume that all keys are distinct and that the compact list is also sorted, that is, key[i] < key[next[i]] for all i = 1, 2, …, n such that next[i] ≠ nil. We will also assume that we have a variable L that contains the index of the first element on the list. Under these assumptions, you will show that we can use the following randomized algorithm to search the list in O(\(\sqrt{n}\)) expected time.
CompactListSearch(L, n, k)
 i = L
 while i ≠ nil and key[i] < k
 j = Random(1, n)
 if key[i] < key[j] and key[j] ≤ k
 i = j
 if key[i] == k
 return i
 i = next[i]
 if i == nil or key[i] > k
 return nil
 else return i
If we ignore lines 3–7 of the procedure, we have an ordinary algorithm for searching a sorted linked list, in which index i points to each position of the list in turn. The search terminates once the index i “falls off” the end of the list or once key[i] ≥ k. In the latter case, if key[i] = k, clearly we have found a key with the value k. If, however, key[i] > k, then we will never find a key with the value k, and so terminating the search was the right thing to do.
Lines 3–7 attempt to skip ahead to a randomly chosen position j. Such a skip benefits us if key[j] is larger than key[i] and no larger than k; in such a case, j marks a position in the list that i would have to reach during an ordinary list search. Because the list is compact, we know that any choice of j between 1 and n indexes some object in the list rather than a slot on the free list.
Instead of analyzing the performance of CompactListSearch directly, we shall analyze a related algorithm, CompactListSearch′, which executes two separate loops. This algorithm takes an additional parameter t which determines an upper bound on the number of iterations of the first loop.
CompactListSearch′(L, n, k, t)
 i = L
 for q = 1 to t
 j = Random(1, n)
 if key[i] < key[j] and key[j] ≤ k
 i = j
 if key[i] == k
 return i
 while i ≠ nil and key[i] < k
 i = next[i]
 if i == nil or key[i] > k
 return nil
 else return i
To compare the execution of the algorithms CompactListSearch(L, n, k) and CompactListSearch′(L, n, k, t), assume that the sequence of integers returned by the calls of Random(1, n) is the same for both algorithms.
 a. Suppose that CompactListSearch(L, n, k) takes t iterations of the while loop of lines 2–8. Argue that CompactListSearch′(L, n, k, t) returns the same answer and that the total number of iterations of both the for and while loops within CompactListSearch′ is at least t.
In the call CompactListSearch′(L, n, k, t), let \(X_t\) be the random variable that describes the distance in the linked list (that is, through the chain of next pointers) from position i to the desired key k after t iterations of the for loop of lines 2–7 have occurred.
 b. Argue that the expected running time of CompactListSearch′(L, n, k, t) is O(t + E[\(X_t\)]).
 c. Show that E[\(X_t\)] ≤ \(∑_{r = 1}^n (1  r / n)^t\). (Hint: Use equation (C.25).)
 d. Show that \(∑_{r = 0}^{n  1} r^t ≤ n^{t + 1} / (t + 1)\).
 e. Prove that E[\(X_t\)] ≤ n / (t + 1).
 f. Show that CompactListSearch′(L, n, k, t) runs in O(t + n / t) expected time.
 g. Conclude that CompactListSearch runs in O(\(\sqrt{n}\)) expected time.
 h. Why do we assume that all keys are distinct in CompactListSearch? Argue that random skips do not necessarily help asymptotically when the list contains repeated key values.
Skipped.
11 Hash Tables
11.1 Directaddress tables
11.11
Suppose that a dynamic set S is represented by a directaddress table T of length m. Describe a procedure that finds the maximum element of S. What is the worstcase performance of your procedure?
Θ(m).
11.12
A bit vector is simply an array of bits (0s and 1s). A bit vector of length m takes much less space than an array of m pointers. Describe how to use a bit vector to represent a dynamic set of distinct elements with no satellite data. Dictionary operations should run in O(1) time.
We use the value of ith bit to indicate whether the key i exists, then the implementation is trivial.
11.13
Suggest how to implement a directaddress table in which the keys of stored elements do not need to be distinct and the elements can have satellite data. All three dictionary operations (Insert, Delete, and Search) should run in O(1) time. (Don’t forget that Delete takes as an argument a pointer to an object to be deleted, not a key.)
For all elements with key k, store them in a doubly linked list, then store the doubly linked list in position i of the directaddress table.
11.14 ★
We wish to implement a dictionary by using direct addressing on a huge array. At the start, the array entries may contain garbage, and initializing the entire array is impractical because of its size. Describe a scheme for implementing a directaddress dictionary on a huge array. Each stored object should use O(1) space; the operations Search, Insert, and Delete should take O(1) time each; and initializing the data structure should take O(1) time. (Hint: Use an additional array, treated somewhat like a stack whose size is the number of keys actually stored in the dictionary, to help determine whether a given entry in the huge array is valid or not.)
Solution is implemented here.
11.2 Hash tables
11.21
Suppose we use a hash function h to hash n distinct keys into an array T of length m. Assuming simple uniform hashing, what is the expected number of collisions? More precisely, what is the expected cardinality of {{k, l} : k ≠ l and h(k) = h(l)}?
Let \(X_i\) be the random variable that represents the number of elements in the ith position, then the expected number of collisions is:
E[\(∑_{i = 1}^m X_i \left(X_i  1\right) / 2\)]
= \(∑_{i = 1}^m \operatorname{E} \left[X_i \left(X_i  1\right) / 2\right]\)
= m E[\(X_1 \left(X_1  1\right) / 2\)]
= m (E[\(X_1^2\)]  E[\(X_1\)]) / 2
= m (E[\(X_1^2\)]  n / m) / 2
= m (\(\operatorname{E}^2 \left[X_i\right]\) + Var[\(X_i\)]  n / m) / 2
= m (\((n / m)^2\) + n (1 / m) (m  1) / m  n / m) / 2
= n (n  1) / (2 m).
That is way too complicated.
Another way:
Let \(X_i\) be the ith key, then the expected number of collisions is:
E[\(∑_{i = 1}^{n  1} ∑_{j = i + 1}^n \Pr\left\lbrace h\left(X_i\right) = h\left(X_j\right) \right\rbrace\)]
= E[\(∑_{i = 1}^{n  1} ∑_{j = i + 1}^n 1 / m\)]
= \(∑_{i = 1}^{n  1} ∑_{j = i + 1}^n 1 / m\)
= n (n  1) / (2 m).
11.22
Demonstrate what happens when we insert the keys 5, 28, 19, 15, 20, 33, 12, 17, 10 into a hash table with collisions resolved by chaining. Let the table have 9 slots, and let the hash function be h(k) = k mod 9.
Skipped.
11.23
Professor Marley hypothesizes that he can obtain substantial performance gains by modifying the chaining scheme to keep each list in sorted order. How does the professor’s modification affect the running time for successful searches, unsuccessful searches, insertions, and deletions?
 Successful searches: Worst case running time decreases to Θ(lg n).
 Unsuccessful searches: Worst case running time decreases to Θ(lg n).
 Insertions: Worst case running time increases to Θ(n).
 Deletions: Unaffected.
11.24
Suggest how to allocate and deallocate storage for elements within the hash table itself by linking all unused slots into a free list. Assume that one slot can store a flag and either one element plus a pointer or two pointers. All dictionary and freelist operations should run in O(1) expected time. Does the free list need to be doubly linked, or does a singly linked free list suffice?
Solution is implemented here.
11.25
Suppose that we are storing a set of n keys into a hash table of size m. Show that if the keys are drawn from a universe U with U > n m, then U has a subset of size n consisting of keys that all hash to the same slot, so that the worstcase searching time for hashing with chaining is Θ(n).
This is like putting n m balls into m buckets, there must be at least one bucket contains more than or equal to n balls.
11.26
Suppose we have stored n keys in a hash table of size m, with collisions resolved by chaining, and that we know the length of each chain, including the length L of the longest chain. Describe a procedure that selects a key uniformly at random from among the keys in the hash table and returns it in expected time O(L ⋅ (1 + 1 / α)).
First, randomly generate an integer i within [1, m], then generate an random integer j within [1, L], if the length of the ith chain is less than or equal to j, return the jth element of the ith chain as the result, otherwise repeat the process.
Rationalization:
j
↑
L │ ▌
│ ▌ ▌ ▌
│ ▌▌ ▌ ▌ ▌
│▌▌▌ ▌▌ ▌ ▌ ▌ ▌ ▌ ▌
│▌▌▌ ▌▌▌ ▌ ▌▌▌ ▌▌▌▌▌ ▌▌
──┼───────────────────────→ i
0 │ m
Time complexity analysis:
The process is a sequence of Bernoulli trials with a probability p = n / (m L). So the number of iterations needed to find an element follows geometric distribution, so the expected number of iterations for finding the position of an element is 1 / p = m L / n = L / α. But it is not done, we still need to get the value of the selected element in a linked list, which takes O(L) time. So the total expected running time is Θ(L / α) + O(L) = O(L ⋅ (1 + 1 / α)).
11.3 Hash functions
11.31
Suppose we wish to search a linked list of length n, where each element contains a key k along with a hash value h(k). Each key is a long character string. How might we take advantage of the hash values when searching the list for an element with a given key?
Precompute the hash value of the given key. Then when searching the linked list, first compare the hash value, only if the hash values are equal, we compare the actual string.
11.32
Suppose that we hash a string of r characters into m slots by treating it as a radix128 number and then using the division method. We can easily represent the number m as a 32bit computer word, but the string of r characters, treated as a radix128 number, takes many words. How can we apply the division method to compute the hash value of the character string without using more than a constant number of words of storage outside the string itself?
Solution is implemented here.
11.33
Consider a version of the division method in which h(k) = k mod m, where m = \(2^p\)  1 and k is a character string interpreted in radix \(2^p\). Show that if we can derive string x from string y by permuting its characters, then x and y hash to the same value. Give an example of an application in which this property would be undesirable in a hash function.
According to the solution to exercise 11.32, we can calculate the hash value using a loop whose body is:
result = (result × \(2^p\) + S[i]) mod m.
If m = \(2^p\)  1, the body become:
result
= (result × \(2^p\) + S[i]) mod (\(2^p\)  1)
= ((result mod (\(2^p\)  1)) × (\(2^p\) mod (\(2^p\)  1)) + S[i] mod (\(2^p\)  1)) mod (\(2^p\)  1)
= ((result mod (\(2^p\)  1)) × 1 + S[i] mod (\(2^p\)  1)) mod (\(2^p\)  1)
= ((result mod (\(2^p\)  1)) + S[i] mod (\(2^p\)  1)) mod (\(2^p\)  1)
= (result + S[i] mod (\(2^p\)  1)) mod (\(2^p\)  1).
Where S[i] is the ith element of the string.
So the final result is just \(\left(∑_{i = 1}^n S[i] \bmod m\right) \bmod m\). Notice that the order of S[i] does not matter in calculating the hash value.
As for the example, consider storing the anagrams of a word in a hash table.
11.34
Consider a hash table of size m = 1000 and a corresponding hash function h(k) = ⌊m (k A mod 1)⌋ for A = (\(\sqrt{5}\)  1) / 2. Compute the locations to which the keys 61, 62, 63, 64, and 65 are mapped.
Skipped.
11.35 ★
Define a family ℋ of hash functions from a finite set U to a finite set B to be ϵuniversal if for all pairs of distinct elements k and l in U,
Pr{h(k) = h(l)} ≤ ϵ,
where the probability is over the choice of the hash function h drawn at random from the family ℋ. Show that an ϵuniversal family of hash functions must have
ϵ ≥ \(\displaystyle \frac{1}{\leftB\right}  \frac{1}{\leftU\right}\).
Skipped.
11.36 ★
Let U be the set of ntuples of values drawn from \(ℤ_p\), and let B = \(ℤ_p\), where p is prime. Define the hash function \(h_b\): U → B for b ∈ \(ℤ_p\) on an input ntuple ⟨\(a_0\), \(a_1\), …, \(a_{n  1}\)⟩ from U as
\(h_b\left(⟨a_0, a_1, …, a_{n  1}⟩\right) = \left(\displaystyle ∑_{j = 0}^{n = 1} a_j b^j\right) \bmod p\).
and let ℋ = {\(h_b\) : b ∈ \(ℤ_p\)}. Argue that ℋ is ((n  1) / p)universal according to the definition of ϵuniversal in Exercise 11.35. (Hint: See Exercise 31.44.)
Skipped.
11.4 Open addressing
11.41
Consider inserting the keys 10, 22, 31, 4, 15, 28, 17, 88, 59 into a hash table of length m = 11 using open addressing with the auxiliary hash function h′(k) = k. Illustrate the result of inserting these keys using linear probing, using quadratic probing with \(c_1\) = 1 and \(c_2\) = 3, and using double hashing with \(h_1\)(k) = k and \(h_2\)(k) = 1 + (k mod (m  1)).
Skipped.
11.42
Write pseudocode for HashDelete as outlined in the text, and modify HashInsert to handle the special value deleted.
Skipped.
11.43
Consider an openaddress hash table with uniform hashing. Give upper bounds on the expected number of probes in an unsuccessful search and on the expected number of probes in a successful search when the load factor is 3 / 4 and when it is 7 / 8.
Skipped.
11.44 ★
Suppose that we use double hashing to resolve collisions—that is, we use the hash function h(k, i) = (\(h_1\)(k) + i \(h_2\)(k)) mod m. Show that if m and \(h_2\)(k) have greatest common divisor d ≥ 1 for some key k, then an unsuccessful search for key k examines (1 / d)th of the hash table before returning to slot h_1(k). Thus, when d = 1, so that m and \(h_2\)(k) are relatively prime, the search may examine the entire hash table. (Hint: See Chapter 31.)
Skipped.
11.45 ★
Consider an openaddress hash table with a load factor α. Find the nonzero value α for which the expected number of probes in an unsuccessful search equals twice the expected number of probes in a successful search. Use the upper bounds given by Theorems 11.6 and 11.8 for these expected numbers of probes.
Skipped.
11.5 Perfect hashing ★
11.51 ★
Suppose that we insert n keys into a hash table of size m using open addressing and uniform hashing. Let p(n, m) be the probability that no collisions occur. Show that p(n, m) ≤ \(e^{n (n  1) / 2 m}\). (Hint: See equation (3.12).) Argue that when n exceeds \(\sqrt{m}\), the probability of avoiding collisions goes rapidly to zero.
p(n, m)
= \(\frac{P(m, n)}{m^n}\)
Skipped.
11.X Problems
111 Longestprobe bound for hashing
Suppose that we use an openaddressed hash table of size m to store n ≤ m / 2 items.
a. Assuming uniform hashing, show that for i = 1, 2, …, n, the probability is at most \(2^{k}\) that the ith insertion requires strictly more than k probes.
b. Show that for i = 1, 2, …, n, the probability is O(1 / \(n^2\)) that the ith insertion requires more than 2 lg n probes.
Let the random variable \(X_i\) denote the number of probes required by the ith insertion. You have shown in part (b) that Pr{\(X_i\) > 2 lg n} = O(1 / \(n^2\)). Let the random variable X = \(\max_{1 ≤ i ≤ n} X_i\) denote the maximum number of probes required by any of the n insertions.
c. Show that Pr{X > 2 lg n} = O(1 / n).
d. Show that the expected length E[X] of the longest probe sequence is O(lg n).
Skipped.
112 Slotsize bound for chaining
Suppose that we have a hash table with n slots, with collisions resolved by chaining, and suppose that n keys are inserted into the table. Each key is equally likely to be hashed to each slot. Let M be the maximum number of keys in any slot after all the keys have been inserted. Your mission is to prove an O(lg n / lg lg n) upper bound on E[M], the expected value of M.
a. Argue that the probability \(Q_k\) that exactly k keys hash to a particular slot is given by \(\displaystyle Q_k = \left(\frac{1}{n}\right)^k \left(1  \frac{1}{n}\right)^{n  k} \binom{n}{k}\).
b. Let \(P_k\) be the probability that M = k, that is, the probability that the slot containing the most keys contains k keys. Show that \(P_k\) ≤ n \(Q_k\).
c. Use Stirling’s approximation, equation (3.18), to show that \(Q_k < e^k / k^k\).
d. Show that there exists a constant c > 1 such that \(Q_{k_0}\) < 1 / \(n^3\) for \(k_0\) = c lg n / lg lg n. Conclude that \(P_k\) < 1 / \(n^2\) for k ≥ \(k_0\) = c lg n / lg lg n.
e. Argue that
E[M] ≤ \(\displaystyle\Pr\left\lbrace M > \frac{c \lg n}{\lg \lg n} \right\rbrace ⋅ n + \Pr\left\lbrace M ≤ \frac{c \lg n}{\lg \lg n} \right\rbrace ⋅ \frac{c \lg n}{\lg \lg n}\).
Conclude that E[M] = O(lg n / lg lg n).
Skipped.
113 Quadratic probing
Suppose that we are given a key k to search for in a hash table with positions 0, 1, …, m  1, and suppose that we have a hash function h mapping the key space into the set {0, 1, …, m  1}. The search scheme is as follows:
 Compute the value j = h(k), and set i = 0.
 Probe in position j for the desired key k. If you find it, or if this position is empty, terminate the search.
 Set i = i + 1. If i now equals m, the table is full, so terminate the search. Otherwise, set j = (i + j) / mod m, and return to step 2.
Assume that m is a power of 2.
a. Show that this scheme is an instance of the general “quadratic probing” scheme by exhibiting the appropriate constants \(c_1\) and \(c_2\) for equation (11.5).
b. Prove that this algorithm examines every table position in the worst case.
Skipped.
114 Hashing and authentication
Let ℋ be a class of hash functions in which each hash function h ∈ ℋ maps the universe U of keys to {0, 1, …, m  1}. We say that ℋ is kuniversal if, for every fixed sequence of k distinct keys \(⟨x^{(1)}, x^{(2)}, …, x^{(k)}⟩\) and for any h chosen at random from ℋ, the sequence \(⟨h(x^{(1)}), h(x^{(2)}), …, h(x^{(k)})⟩\) is equally likely to be any of the \(m^k\) sequences of length k with elements drawn from {0, 1, …, m  1}.
a. Show that if the family ℋ of hash functions is 2universal, then it is universal.
b. Suppose that the universe U is the set of ntuples of values drawn from \(ℤ_p\) = {0, 1, …, p  1}, where p is prime. Consider an element x = \(⟨x_0, x_1, …, x_{n  1}⟩\) ∈ U. For any ntuple a = \(⟨a_0, a_1, …, a_{n  1}⟩\) ∈ U, define the hash function \(h_a\) by
\(\displaystyle h_a(x) = \left(∑_{j = 0}^{n  1} a_j x_j\right) \bmod p\).
Let ℋ = {\(h_a\)}. Show that ℋ is universal, but not 2universal. (Hint: Find a key for which all hash functions in ℋ produce the same value.)
c. Suppose that we modify ℋ slightly from part (b): for any a ∈ U and for any b ∈ \(ℤ_p\), define
\(\displaystyle h'_{a b}(x) = \left(∑_{j = 0}^{n  1} a_j x_j + b\right) \bmod p\)
and ℋ′ = {\(h'_{a b}\)}. Argue that ℋ′ is 2universal. (Hint: Consider fixed ntuples x ∈ U and y ∈ U, with \(x_i\) ≠ \(y_i\) for some i. What happens to \(h'_{a b}(x)\) and \(h'_{a b}(y)\) as \(a_i\) and b range over \(ℤ_p\)?)
d. Suppose that Alice and Bob secretly agree on a hash function h from a 2universal family ℋ of hash functions. Each h ∈ ℋ maps from a universe of keys U to \(ℤ_p\), where p is prime. Later, Alice sends a message m to Bob over the Internet, where m ∈ U. She authenticates this message to Bob by also sending an authentication tag t = h(m), and Bob checks that the pair (m, t) he receives indeed satisfies t = h(m). Suppose that an adversary intercepts (m, t) en route and tries to fool Bob by replacing the pair (m, t) with a different pair (m′, t′). Argue that the probability that the adversary succeeds in fooling Bob into accepting (m′, t′) is at most 1 / p, no matter how much computing power the adversary has, and even if the adversary knows the family ℋ of hash functions used.
Skipped.
12 Binary Search Trees
12.1 What is a binary search tree?
12.11
For the set of {1, 4, 5, 10, 16, 17, 21} of keys, draw binary search trees of heights 2, 3, 4, 5, and 6.
Skipped.
12.12
What is the difference between the binarysearchtree property and the minheap property (see page 153)? Can the minheap property be used to print out the keys of an nnode tree in sorted order in O(n) time? Show how, or explain why not.
Binarysearchtree property requires nodes in left subtree of every node being less than or equal to a node, but minheap property requires nodes in left subtree of every node being greater than or equal to the node.
We can not print out the keys of an nnode tree with minheap property in sorted order in O(n) time. Remember heapsort? If we can do that, we can get a comparison sorting algorithm runs in O(n) time, which is impossible.
12.13
Give a nonrecursive algorithm that performs an inorder tree walk. (Hint: An easy solution uses a stack as an auxiliary data structure. A more complicated, but elegant, solution uses no stack but assumes that we can test two pointers for equality.)
Solution is implemented here.
12.14
Give recursive algorithms that perform preorder and postorder tree walks in Θ(n) time on a tree of n nodes.
Solution is implemented here.
12.15
Argue that since sorting n elements takes Ω(n lg n) time in the worst case in the comparison model, any comparisonbased algorithm for constructing a binary search tree from an arbitrary list of n elements takes Ω(n lg n) time in the worst case.
Like exercise 12.12, If we can construct a binary search tree within o(n lg n) time, we can get a comparison sorting algorithm that takes o(n lg n) time, which is impossible.
12.2 Querying a binary search tree
12.21
Suppose that we have numbers between 1 and 1000 in a binary search tree, and we want to search for the number 363. Which of the following sequences could not be the sequence of nodes examined?
 a. 2, 252, 401, 398, 330, 344, 397, 363.
 b. 924, 220, 911, 244, 898, 258, 362, 363.
 c. 925, 202, 911, 240, 912, 245, 363.
 d. 2, 399, 387, 219, 266, 382, 381, 278, 363.
 e. 935, 278, 347, 621, 299, 392, 358, 363.
c and e are not valid sequences.
Here is an algorithm for detecting valid sequences.
12.22
Write recursive versions of TreeMinimum and TreeMaximum.
Solution is implemented here.
12.23
Write the TreePredecessor procedure.
Solution is implemented here.
12.24
Professor Bunyan thinks he has discovered a remarkable property of binary search trees. Suppose that the search for key k in a binary search tree ends up in a leaf. Consider three sets: A, the keys to the left of the search path; B, the keys on the search path; and C, the keys to the right of the search path. Professor Bunyan claims that any three keys a ∈ A, b ∈ B, and c ∈ C must satisfy a ≤ b ≤ c. Give a smallest possible counterexample to the professor’s claim.
┌───┐
│ 1 │
└─┬─┘
┌───┴───┐
┌──┴──┐ ┌─┴─┐
│ nil │ │ 3 │
└─────┘ └─┬─┘
┌──┴───┐
┌─┴─┐ ┌─┴─┐
│ 2 │ │ 4 │
└───┘ └───┘
Searching for 4, we have 2 ∈ A and 1 ∈ B.
12.25
Show that if a node in a binary search tree has two children, then its successor has no left child and its predecessor has no right child.
If a node has left child then its predecessor is the right most node of the left child, if the right most node has right child, then it is not the right most child, by contradiction, we know the predecessor has no right child.
The same applies to the successor.
12.26
Consider a binary search tree T whose keys are distinct. Show that if the right subtree of a node x in T is empty and x has a successor y, then y is the lowest ancestor of x whose left child is also an ancestor of x. (Recall that every node is its own ancestor.)
Skipped.
12.27
An alternative method of performing an inorder tree walk of an nnode binary search tree finds the minimum element in the tree by calling TreeMinimum and then making n  1 calls to TreeSuccessor. Prove that this algorithm runs in Θ(n) time.
Skipped.
12.28
Prove that no matter what node we start at in a heighth binary search tree, k successive calls to TreeSuccessor take O(k + h) time.
Skipped.
12.29
Let T be a binary search tree whose keys are distinct, let x be a leaf node, and let y be its parent. Show that y.key is either the smallest key in T larger than x.key or the largest key in T smaller than x.key.
By the definition of Successor, we know that:
 If x is the left child of y, then y is the successor of x, so y.key is the smallest key in T larger than x.key.
 Otherwise, x is the right child of y, then x is the successor of y, so y.key is the largest key in T smaller than x.key.
12.3 Insertion and deletion
12.31
Give a recursive version of the TreeInsert procedure.
Solution is implemented here.
12.32
Suppose that we construct a binary search tree by repeatedly inserting distinct values into the tree. Argue that the number of nodes examined in searching for a value in the tree is one plus the number of nodes examined when the value was first inserted into the tree.
Note that the new value is only inserted to the leaf position, so inserting a new value does not affect the relative positions of the old values, which means once a node is inserted into the tree, it stays at its position all the time.
The number of nodes to be examined when the node was inserted into the tree equals to the number of ancestors except the node itself, and searching examines all the node’s ancestors. So the claim is true.
12.33
We can sort a given set of n numbers by first building a binary search tree containing these numbers (using TreeInsert repeatedly to insert the numbers one by one) and then printing the numbers by an inorder tree walk. What are the worstcase and bestcase running times for this sorting algorithm?
Worstcase running time is Θ(\(n^2\)) when the numbers are already sorted.
Bestcase running time is Θ(n log n) when the tree is a complete binary tree.
12.34
Is the operation of deletion “commutative” in the sense that deleting x and then y from a binary search tree leaves the same tree as deleting y and then x? Argue why it is or give a counterexample.
Deletion is not “commutative”.
Counterexample: deleting 1 and 2 from the following tree:
┌───┐
│ 2 │
└─┬─┘
┌──┴───┐
┌─┴─┐ ┌─┴─┐
│ 1 │ │ 4 │
└───┘ └─┬─┘
┌───┴───┐
┌─┴─┐ ┌──┴──┐
│ 3 │ │ nil │
└───┘ └─────┘
Delete 1 first, then delete 2, we will get the following tree:
┌───┐
│ 4 │
└─┬─┘
┌───┴───┐
┌─┴─┐ ┌──┴──┐
│ 3 │ │ nil │
└───┘ └─────┘
Delete 2 first, then delete 1, we will get the following tree:
┌───┐
│ 3 │
└─┬─┘
┌───┴───┐
┌──┴──┐ ┌─┴─┐
│ nil │ │ 4 │
└─────┘ └───┘
12.35
Suppose that instead of each node x keeping the attribute x.p, pointing to x’s parent, it keeps x.succ, pointing to x’s successor. Give pseudocode for Search, Insert, and Delete on a binary search tree T using this representation. These procedures should operate in time O(h), where h is the height of the tree T. (Hint: You may wish to implement a subroutine that returns the parent of a node.)
Skipped.
12.36
When node z in TreeDelete has two children, we could choose node y as its predecessor rather than its successor. What other changes to TreeDelete would be necessary if we did so? Some have argued that a fair strategy, giving equal priority to predecessor and successor, yields better empirical performance. How might TreeDelete be changed to implement such a fair strategy?
Skipped.
12.4 Randomly built binary search trees ★
12.41
Prove equation (12.3).
Proof by induction:

Base case: If n = 1,
\(∑_{i = 0}^{n  1} \binom{i + 3}{3}\) = \(∑_{i = 0}^{0} \binom{i + 3}{3}\) = \(\binom{3}{3}\) = 1.
\(\binom{n + 3}{4}\) = \(\binom{4}{4}\) = 1.
So the claim holds.

Inductive cases: If n > 1,
\(∑_{i = 0}^{n  1} \binom{i + 3}{3}\)
= \(∑_{i = 0}^{n  2} \binom{i + 3}{3} + \binom{n + 2}{3}\)
= \(\binom{n + 2}{4} + \binom{n + 2}{3}\)
= \(\frac{(n + 2)!}{4! (n  2)!} + \frac{(n + 2)!}{3! (n  1)!}\)
= \(\frac{(n + 2)!}{(n  1)!} \left(\frac{n  1}{4!} + \frac{1}{3!}\right)\)
= \(\frac{(n + 2)!}{(n  1)!} \frac{n + 3}{4!}\)
= \(\frac{(n + 3)!}{(n  1)! 4!}\)
= \(\binom{n + 3}{4}\).
12.42
Describe a binary search tree on n nodes such that the average depth of a node in the tree is Θ(lg n) but the height of the tree is ω(lg n). Give an asymptotic upper bound on the height of an nnode binary search tree in which the average depth of a node is Θ(lg n).
Skipped.
12.43
Show that the notion of a randomly chosen binary search tree on n keys, where each binary search tree of n keys is equally likely to be chosen, is different from the notion of a randomly built binary search tree given in this section. (Hint: List the possibilities when n = 3.)
Skipped.
12.44
Show that the function f(x) = \(2^x\) is convex.
Skipped.
12.45 ★
Consider RandomizedQuicksort operating on a sequence of n distinct input numbers. Prove that for any constant k > 0, all but O(1 / \(n^k\)) of the n! input permutations yield an O(n lg n) running time.
Skipped.
12.X Problems
121 Binary search trees with equal keys
Equal keys pose a problem for the implementation of binary search trees.
 a. What is the asymptotic performance of TreeInsert when used to insert n items with identical keys into an initially empty binary search tree?
We propose to improve TreeInsert by testing before line 5 to determine whether z.key = x.key and by testing before line 11 to determine whether z.key = y.key.
If equality holds, we implement one of the following strategies. For each strategy, find the asymptotic performance of inserting n items with identical keys into an initially empty binary search tree. (The strategies are described for line 5, in which we compare the keys of z and x. Substitute y for x to arrive at the strategies for line 11.)
 b. Keep a boolean flag x.b at node x, and set x to either x.left or x.right based on the value of x.b, which alternates between false and true each time we visit x while inserting a node with the same key as x.
 c. Keep a list of nodes with equal keys at x, and insert z into the list.
 d. Randomly set x to either x.left or x.right. (Give the worstcase performance and informally derive the expected running time.)
 a. Θ(\(n^2\)).
 b. Θ(n lg n).
 c. Θ(n lg n).
 d. Θ(n).
 d. Worstcase running time is Θ(\(n^2\)), expected running time is Θ(n lg n).
122 Radix trees
Given two strings a = \(a_0 a_1 … a_p\) and b = \(b_0 b_1 … b_q\), where each \(a_i\) and each \(b_j\) is in some ordered set of characters, we say that string a is lexicographically less than string b if either
 there exists an integer j, where 0 ≤ j ≤ min(p, q), such that \(a_i\) = \(b_i\) for all i = 0, 1, …, j  1 and \(a_j\) < \(b_j\), or
 p < q and \(a_i\) = \(b_i\) for all i = 0, 1, …, p.
For example, if a and b are bit strings, then 10100 < 10110 by rule 1 (letting j = 3) and 10100 < 101000 by rule 2. This ordering is similar to that used in Englishlanguage dictionaries.
The radix tree data structure shown in Figure 12.5 stores the bit strings 1011, 10, 011, 100, and 0. When searching for a key a = \(a_0 a_1 … a_p\), we go left at a node of depth i if \(a_i\) = 0 and right if \(a_i\) = 1. Let S be a set of distinct bit strings whose lengths sum to n. Show how to use a radix tree to sort S lexicographically in Θ(n) time. For the example in Figure 12.5, the output of the sort should be the sequence 0, 011, 10, 100, 1011.
┌──┐ │//│ └┬─┘ ┌──────────┴──────────┐ 0 1 ┌─┴─┐ ┌┴─┐ │ 0 │ │//│ └─┬─┘ └┬─┘ └──┐ ┌──┘ 1 0 ┌┴─┐ ┌─┴──┐ │//│ │ 10 │ └┬─┘ └─┬──┘ └───┐ ┌──┴───┐ 1 0 1 ┌──┴──┐ ┌──┴──┐ ┌┴─┐ │ 011 │ │ 100 │ │//│ └─────┘ └─────┘ └┬─┘ └───┐ 1 ┌──┴───┐ │ 1011 │ └──────┘
Figure 12.5 A radix tree storing the bit strings 1011, 10, 011, 100, and 0. We can determine each node’s key by traversing the simple path from the root to that node. There is no need, therefore, to store the keys in the nodes; the keys appear here for illustrative purposes only. Nodes are heavily shaded if the keys corresponding to them are not in the tree; such nodes are present only to establish a path to other nodes.
Solution is implemented here.
123 Average node depth in a randomly built binary search tree
In this problem, we prove that the average depth of a node in a randomly built binary search tree with n nodes is O(lg n). Although this result is weaker than that of Theorem 12.4, the technique we shall use reveals a surprising similarity between the building of a binary search tree and the execution of RandomizedQuicksort from Section 7.3.
We define the total path length P(T) of a binary tree T as the sum, over all nodes x in T, of the depth of node x, which we denote by d(x, T).
a. Argue that the average depth of a node in T is
\(\displaystyle \frac{1}{n} ∑_{x ∈ T} d(x, T) = \frac{1}{n} P(T)\).
Thus, we wish to show that the expected value of P(T) is O(n lg n).
b. Let \(T_L\) and \(T_R\) denote the left and right subtrees of tree T, respectively. Argue that if T has n nodes, then
P(T) = P(\(T_L\)) + P(\(T_R\)) + n  1.
c. Let P(n) denote the average total path length of a randomly built binary search tree with n nodes. Show that
\(\displaystyle P(n) = \frac{1}{n} ∑_{i = 0}^{n  1}(P(i) + P(n  i  1) + n  1)\).
d. Show how to rewrite P(n) as
\(\displaystyle P(n) = \frac{2}{n} ∑_{k = 1}^{n  1} P(k) + Θ(n)\).
e. Recalling the alternative analysis of the randomized version of quicksort given in Problem 73, conclude that P(n) = O(n lg n).
At each recursive invocation of quicksort, we choose a random pivot element to partition the set of elements being sorted. Each node of a binary search tree partitions the set of elements that fall into the subtree rooted at that node.
 f. Describe an implementation of quicksort in which the comparisons to sort a set of elements are exactly the same as the comparisons to insert the elements into a binary search tree. (The order in which comparisons are made may differ, but the same comparisons must occur.)
Skipped.
124 Number of different binary trees
Let \(b_n\) denote the number of different binary trees with n nodes. In this problem, you will find a formula for \(b_n\), as well as an asymptotic estimate.
a. Show that \(b_0\) = 1 and that, for n ≥ 1,
\(\displaystyle b_n = ∑_{k = 0}^{n  1} b_k b_{n  1  k}\).
b. Referring to Problem 44 for the definition of a generating function, let B(x) be the generating function
B(x) = \(\displaystyle ∑_{n = 0}^∞ b_n x^n\).
Show that B(x) = x \(B(x)^2\) + 1, and hence one way to express B(x) in closed form is
B(x) = \(\dfrac{1}{2 x} \left(1  \sqrt{1  4 x}\right)\).
The Taylor expansion of f(x) around the point x = a is given by
f(x) = \(\displaystyle ∑_{k = 0}^∞ \frac{f^{(k)}(a)}{k!}(x  a)^k\).
where \(f^{(k)}(x)\) is the kth derivative of f evaluated at x.
c. Show that
\(\displaystyle b_n = \frac{1}{n + 1} \binom{2 n}{n}\)
(the nth Catalan number) by using the Taylor expansion of \(\sqrt{1  4 x}\) around x = 0. (If you wish, instead of using the Taylor expansion, you may use the generalization of the binomial expansion (C.4) to nonintegral exponents n, where for any real number n and for any integer k, we interpret \(\binom{n}{k}\) to be n (n  1) ⋯ (n  k + 1) / k! if k ≥ 0, and 0 otherwise.)
d. Show that
\(b_n = \dfrac{4^n}{\sqrt{\pi} n^{3 / 2}}\) (1 + O(1 / n)).

a
 There is only one type of binary tree of zero nodes, that is, no tree, so \(b_0\) = 1.
 For n > 0, if the left child of the root has k nodes, then the right child has n  1  k nodes, for this situation, there can be \(b_k b_{n  1  k}\) different kinds of trees. Since k must be in range [0, n  1], the total number of kinds of binary trees is \(∑_{k = 0}^{n  1} b_k b_{n  1  k}\).

b
Skipped.

c
Skipped.

d
Skipped.
13 RedBlack Trees
13.1 Properties of redblack trees
13.11
In the style of Figure 13.1(a), draw the complete binary search tree of height 3 on the keys {1, 2, …, 15}. Add the nil leaves and color the nodes in three different ways such that the blackheights of the resulting redblack trees are 2, 3, and 4.
There you go:
┌─────┐
│ [8] │
└──┬──┘
┌─────────────┴──────────────┐
┌─┴─┐ ┌─┴──┐
│ 4 │ │ 12 │
└─┬─┘ └─┬──┘
┌──────┴──────┐ ┌───────┴───────┐
┌──┴──┐ ┌──┴──┐ ┌──┴───┐ ┌──┴───┐
│ [2] │ │ [6] │ │ [10] │ │ [14] │
└──┬──┘ └──┬──┘ └──┬───┘ └──┬───┘
┌──┴───┐ ┌──┴───┐ ┌──┴───┐ ┌───┴───┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴──┐ ┌─┴──┐ ┌─┴──┐
│ 1 │ │ 3 │ │ 5 │ │ 7 │ │ 9 │ │ 11 │ │ 13 │ │ 15 │
└───┘ └───┘ └───┘ └───┘ └───┘ └────┘ └────┘ └────┘
┌─────┐
│ [8] │
└──┬──┘
┌─────────────┴──────────────┐
┌──┴──┐ ┌──┴───┐
│ [4] │ │ [12] │
└──┬──┘ └──┬───┘
┌──────┴──────┐ ┌───────┴───────┐
┌──┴──┐ ┌──┴──┐ ┌──┴───┐ ┌──┴───┐
│ [2] │ │ [6] │ │ [10] │ │ [14] │
└──┬──┘ └──┬──┘ └──┬───┘ └──┬───┘
┌──┴───┐ ┌──┴───┐ ┌──┴───┐ ┌───┴───┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴──┐ ┌─┴──┐ ┌─┴──┐
│ 1 │ │ 3 │ │ 5 │ │ 7 │ │ 9 │ │ 11 │ │ 13 │ │ 15 │
└───┘ └───┘ └───┘ └───┘ └───┘ └────┘ └────┘ └────┘
┌─────┐
│ [8] │
└──┬──┘
┌─────────────────┴──────────────────┐
┌──┴──┐ ┌──┴───┐
│ [4] │ │ [12] │
└──┬──┘ └──┬───┘
┌────────┴────────┐ ┌─────────┴─────────┐
┌──┴──┐ ┌──┴──┐ ┌──┴───┐ ┌──┴───┐
│ [2] │ │ [6] │ │ [10] │ │ [14] │
└──┬──┘ └──┬──┘ └──┬───┘ └──┬───┘
┌───┴────┐ ┌───┴────┐ ┌───┴────┐ ┌────┴────┐
┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴───┐ ┌──┴───┐ ┌──┴───┐
│ [1] │ │ [3] │ │ [5] │ │ [7] │ │ [9] │ │ [11] │ │ [13] │ │ [15] │
└─────┘ └─────┘ └─────┘ └─────┘ └─────┘ └──────┘ └──────┘ └──────┘
13.12
Draw the redblack tree that results after TreeInsert is called on the tree in Figure 13.1 with key 36. If the inserted node is colored red, is the resulting tree a redblack tree? What if it is colored black?
Skipped.
13.13
Let us define a relaxed redblack tree as a binary search tree that satisfies redblack properties 1, 3, 4, and 5. In other words, the root may be either red or black. Consider a relaxed redblack tree T whose root is red. If we color the root of T black but make no other changes to T, is the resulting tree a redblack tree?
Yes.
13.14
Suppose that we “absorb” every red node in a redblack tree into its black parent, so that the children of the red node become children of the black parent. (Ignore what happens to the keys.) What are the possible degrees of a black node after all its red children are absorbed? What can you say about the depths of the leaves of the resulting tree?
Skipped.
13.15
Show that the longest simple path from a node x in a redblack tree to a descendant leaf has length at most twice that of the shortest simple path from node x to a descendant leaf.
Since all simple paths has the same number of black nodes and the children of a red node must be black, the shortest path will contain no red nodes, and the longest path will interleave between black and red nodes, it is obvious that the longest path has length at most twice that of the shortest path.
13.16
What is the largest possible number of internal nodes in a redblack tree with blackheight k? What is the smallest possible number?
The largest number of internal nodes is \(2^{2 k}\)  1, the smallest possible number is \(2^k\)  1.
13.17
Describe a redblack tree on n keys that realizes the largest possible ratio of red internal nodes to black internal nodes. What is this ratio? What tree has the smallest possible ratio, and what is the ratio?
Skipped.
13.2 Rotations
13.21
Write pseudocode for RightRotate.
I’ll do you one better: Write real code for RightRotate.
13.22
Argue that in every nnode binary search tree, there are exactly n  1 possible rotations.
If a node is the left child of another node, we can do a right rotation on its parent; if a node is the right child of another node, we can do a left rotation on its parent. Since there are n  1 nodes with a parent, we can do n  1 possible rotations.
13.23
Let a, b, and c be arbitrary nodes in subtrees α, β, and γ, respectively, in the right tree of Figure 13.2. How do the depths of a, b, and c change when a left rotation is performed on node x in the figure?
┌───┐ │ 7 │ └─┬─┘ ┌──────┴───────┐ ┌─┴─┐ ┌───┴────┐ │ 4 │ │ 11 (x) │ └─┬─┘ └───┬────┘ ┌──┴───┐ ┌───┴────┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌───┴────┐ │ 3 │ │ 6 │ │ 9 │ │ 18 (y) │ └─┬─┘ └───┘ └───┘ └───┬────┘ ┌──┘ ┌──────┴──────┐ ┌─┴─┐ ┌─┴──┐ ┌─┴──┐ │ 2 │ │ 14 │ │ 19 │ └───┘ └─┬──┘ └─┬──┘ ┌───┴───┐ └───┐ ┌─┴──┐ ┌─┴──┐ ┌─┴──┐ │ 12 │ │ 17 │ │ 22 │ └────┘ └────┘ └─┬──┘ ┌──┘ ┊ ┌─┴──┐ ┊ │ 20 │ LeftRotate(T, X) ┊ └────┘ ┊ ┊ ⇣ ┌───┐ │ 7 │ └─┬─┘ ┌─────────┴─────────┐ ┌─┴─┐ ┌───┴────┐ │ 4 │ │ 18 (y) │ └─┬─┘ └───┬────┘ ┌──┴───┐ ┌──────┴───────┐ ┌─┴─┐ ┌─┴─┐ ┌───┴────┐ ┌─┴──┐ │ 3 │ │ 6 │ │ 11 (x) │ │ 19 │ └─┬─┘ └───┘ └───┬────┘ └─┬──┘ ┌──┘ ┌──┴───┐ └───┐ ┌─┴─┐ ┌─┴─┐ ┌─┴──┐ ┌─┴──┐ │ 2 │ │ 9 │ │ 14 │ │ 22 │ └───┘ └───┘ └─┬──┘ └─┬──┘ ┌───┴───┐ ┌──┘ ┌─┴──┐ ┌─┴──┐ ┌─┴──┐ │ 12 │ │ 17 │ │ 20 │ └────┘ └────┘ └────┘
Figure 13.3 An example of how the procedure LeftRotate(T, x) modifies a binary search tree. Inorder tree walks of the input tree and the modified tree produce the same listing of key values.
 a increases by 1.
 b stays the same.
 c decreases by 1.
13.24
Show that any arbitrary nnode binary search tree can be transformed into any other arbitrary nnode binary search tree using O(n) rotations. (Hint: First show that at most n  1 right rotations suffice to transform the tree into a rightgoing chain.)
One right rotation operation decrease the total number of left child by exactly one and increase the total number of right child by exactly one. So n  1 right rotations suffice to transform the tree into a tree with at least n  1 right child, which can only be a rightgoing chain. This step takes O(n  1) time.
Since the left rotation and right rotation are symmetric, we can recover the original tree structure from a right going chain using left rotations in a reversed order, which takes O(n  1) time.
So to transform one tree A to another tree B, we can first transform it into a right going chain, then recover the structure of tree B. Both steps takes O(n  1) time, so the total running time is O(n).
13.25 ★
We say that a binary search tree \(T_1\) can be rightconverted to binary search tree \(T_2\) if it is possible to obtain \(T_2\) from \(T_1\) via a series of calls to RightRotate. Give an example of two trees \(T_1\) and \(T_2\) such that \(T_1\) cannot be rightconverted to \(T_2\). Then, show that if a tree \(T_1\) can be rightconverted to \(T_2\), it can be rightconverted using O(\(n^2\)) calls to RightRotate.
If A is a right going chain and B is a left going chain, A cannot be rightconverted into B since A doesn’t have left child nodes.
Skipped.
13.3 Insertion
13.31
In line 16 of RBInsert, we set the color of the newly inserted node z to red. Observe that if we had chosen to set z’s color to black, then property 4 of a redblack tree would not be violated. Why didn’t we choose to set z’s color to black?
For any valid redblack tree, property 4 is not violated, then after inserting a black node, red nodes are not affected, so property 4 will still be valid.
But setting z’s color to black will violate property 5, which seems harder to fix than property 4.
13.32
Show the redblack trees that result after successively inserting the keys 41, 38, 31, 12, 19, 8 into an initially empty redblack tree.
Skipped.
13.33
Suppose that the blackheight of each of the subtrees α, β, γ, δ, ε in Figures 13.5 and 13.6 is k. Label each node in each figure with its blackheight to verify that the indicated transformation preserves property 5.
Skipped.
13.34
Professor Teach is concerned that RBInsertFixup might set T.nil.color to red, in which case the test in line 1 would not cause the loop to terminate when z is the root. Show that the professor’s concern is unfounded by arguing that RBInsertFixup never sets T.nil.color to red.
The only possible setting nodes’ color to red is line 7 and line 13, and in both cases, we are sure that z are z.p.p’s grandchild, so z.p.p can’t be T.nil. So RBInsertFixup never sets T.nil.color to red.
13.35
Consider a redblack tree formed by inserting n nodes with RBInsert. Argue that if n > 1, the tree has at least one red node.
Since n > 1, the tree will have a root, so the newly inserted node will have a parent.
If the newly inserted node is a child of a black node, the tree will not need any fixups. Since the newly inserted node is a red node, the tree has at least one red node.
If the newly inserted node is a child of a red node, we will have to do fixup operations on all three cases.
For case 1 of the fixup, one red node will be left on a branch of the new z of next iteration.
For case 2 and 3 of the fixup, after one iteration of this case, two red nodes will be left in the tree.
13.36
Suggest how to implement RBInsert efficiently if the representation for redblack trees includes no storage for parent pointers.
I can do it recursively.
13.4 Deletion
13.41
Argue that after executing RBDeleteFixup, the root of the tree must be black.
Skipped.
13.42
Argue that if in RBDelete both x and x.p are red, then property 4 is restored by the call to RBDeleteFixup(T, x).
Skipped.
13.43
In Exercise 13.32, you found the redblack tree that results from successively inserting the keys 41, 38, 31, 12, 19, 8 into an initially empty tree. Now show the redblack trees that result from the successive deletion of the keys in the order 8, 12, 19, 31, 38, 41.
Skipped.
13.44
In which lines of the code for RBDeleteFixup might we examine or modify the sentinel T.nil?
Skipped.
13.45
In each of the cases of Figure 13.7, give the count of black nodes from the root of the subtree shown to each of the subtrees α, β, …, ζ, and verify that each count remains the same after the transformation. When a node has a color attribute c or c′, use the notation count(c) or count(c′) symbolically in your count.
Skipped.
13.46
Professors Skelton and Baron are concerned that at the start of case 1 of RBDeleteFixup, the node x.p might not be black. If the professors are correct, then lines 5–6 are wrong. Show that x.p must be black at the start of case 1, so that the professors have nothing to worry about.
Because the right child of x.p is red, so the parent must be a black node.
13.47
Suppose that a node x is inserted into a redblack tree with RBInsert and then is immediately deleted with RBDelete. Is the resulting redblack tree the same as the initial redblack tree? Justify your answer.
No. For example:
┌─────┐
│ [2] │
└──┬──┘
┌──┴───┐
┌─┴─┐ ┌─┴─┐
│ 1 │ │ 3 │
└───┘ └───┘
After inserting 4:
┌─────┐
│ [2] │
└──┬──┘
┌───┴────┐
┌──┴──┐ ┌──┴──┐
│ [1] │ │ [3] │
└─────┘ └──┬──┘
┌───┴───┐
┌──┴──┐ ┌─┴─┐
│ nil │ │ 4 │
└─────┘ └───┘
After removing 4:
┌─────┐
│ [2] │
└──┬──┘
┌───┴────┐
┌──┴──┐ ┌──┴──┐
│ [1] │ │ [3] │
└─────┘ └─────┘
13.X Problems
131 Persistent dynamic sets
During the course of an algorithm, we sometimes find that we need to maintain past versions of a dynamic set as it is updated. We call such a set persistent. One way to implement a persistent set is to copy the entire set whenever it is modified, but this approach can slow down a program and also consume much space. Sometimes, we can do much better.
Consider a persistent set S with the operations Insert, Delete, and Search, which we implement using binary search trees as shown in Figure 13.8(a). We maintain a separate root for every version of the set. In order to insert the key 5 into the set, we create a new node with key 5. This node becomes the left child of a new node with key 7, since we cannot modify the existing node with key 7. Similarly, the new node with key 7 becomes the left child of a new node with key 8 whose right child is the existing node with key 10. The new node with key 8 becomes, in turn, the right child of a new root r′ with key 4 whose left child is the existing node with key 3. We thus copy only part of the tree and share some of the nodes with the original tree, as shown in Figure 13.8(b).
Assume that each tree node has the attributes key, left, and right but no parent. (See also Exercise 13.36.)
a. For a general persistent binary search tree, identify the nodes that we need to change to insert a key k or delete a node y.
b. Write a procedure PersistentTreeInsert that, given a persistent tree T and a key k to insert, returns a new persistent tree T′ that is the result of inserting k into T.
c. If the height of the persistent binary search tree T is h, what are the time and space requirements of your implementation of PersistentTreeInsert? (The space requirement is proportional to the number of new nodes allocated.)
d. Suppose that we had included the parent attribute in each node. In this case, PersistentTreeInsert would need to perform additional copying. Prove that PersistentTreeInsert would then require Ω(n) time and space, where n is the number of nodes in the tree.
e. Show how to use redblack trees to guarantee that the worstcase running time and space are O(lg n) per insertion or deletion.
Solutions are implemented here.
132 Join operation on redblack trees
The join operation takes two dynamic sets \(S_1\) and \(S_2\) and an element x such that for any \(x_1\) ∈ \(S_1\) and \(x_2\) ∈ \(S_2\), we have \(x_1\).key ≤ x.key ≤ \(x_2\).key. It returns a set S = \(S_1\) ∪ {x} ∪ \(S_2\). In this problem, we investigate how to implement the join operation on redblack trees.
 a. Given a redblack tree T, let us store its blackheight as the new attribute T.bh. Argue that RBInsert and RBDelete can maintain the bh attribute without requiring extra storage in the nodes of the tree and without increasing the asymptotic running times. Show that while descending through T, we can determine the blackheight of each node we visit in O(1) time per node visited.
We wish to implement the operation RBJoin(\(T_1\), x, \(T_2\)), which destroys \(T_1\) and \(T_2\) and returns a redblack tree T = \(T_1\) ∪ {x} ∪ \(T_2\). Let n be the total number of nodes in \(T_1\) and \(T_2\).
 b. Assume that \(T_1\).bh ≥ \(T_2\).bh. Describe an O(lg n)time algorithm that finds a black node y in \(T_1\) with the largest key from among those nodes whose blackheight is \(T_2\).bh.
 c. Let \(T_y\) be the subtree rooted at y. Describe how \(T_y\) ∪ {x} ∪ \(T_2\) can replace \(T_y\) in O(1) time without destroying the binarysearchtree property.
 d. What color should we make x so that redblack properties 1, 3, and 5 are maintained? Describe how to enforce properties 2 and 4 in O(lg n) time.
 e. Argue that no generality is lost by making the assumption in part (b). Describe the symmetric situation that arises when \(T_1\).bh ≤ \(T_2\).bh.
 f. Argue that the running time of RBJoin is O(lg n).
Skipped.
133 AVL trees
An AVL tree is a binary search tree that is height balanced: for each node x, the heights of the left and right subtrees of x differ by at most 1. To implement an AVL tree, we maintain an extra attribute in each node: x.h is the height of node x. As for any other binary search tree T, we assume that T.root points to the root node.
a. Prove that an AVL tree with n nodes has height O(lg n). (Hint: Prove that an AVL tree of height h has at least \(F_h\) nodes, where \(F_h\) is the hth Fibonacci number.)
b. To insert into an AVL tree, we first place a node into the appropriate place in binary search tree order. Afterward, the tree might no longer be height balanced. Specifically, the heights of the left and right children of some node might differ by 2. Describe a procedure Balance(x), which takes a subtree rooted at x whose left and right children are height balanced and have heights that differ by at most 2, i.e., x.right.h  x.left.h ≤ 2, and alters the subtree rooted at x to be height balanced. (Hint: Use rotations.)
c. Using part (b), describe a recursive procedure AVLInsert(x, z) that takes a node x within an AVL tree and a newly created node z (whose key has already been filled in), and adds z to the subtree rooted at x, maintaining the property that x is the root of an AVL tree. As in TreeInsert from Section 12.3, assume that z.key has already been filled in and that z.left = nil and z.right = nil; also assume that z.h = 0. Thus, to insert the node z into the AVL tree T, we call AVLInsert(T.root, z).
d. Show that AVLInsert, run on an nnode AVL tree, takes O(lg n) time and performs O(1) rotations.
Solution is implemented here.
134 Treaps
If we insert a set of n items into a binary search tree, the resulting tree may be horribly unbalanced, leading to long search times. As we saw in Section 12.4, however, randomly built binary search trees tend to be balanced. Therefore, one strategy that, on average, builds a balanced tree for a fixed set of items would be to randomly permute the items and then insert them in that order into the tree.
What if we do not have all the items at once? If we receive the items one at a time, can we still randomly build a binary search tree out of them?
We will examine a data structure that answers this question in the affirmative. A treap is a binary search tree with a modified way of ordering the nodes. Figure 13.9 shows an example. As usual, each node x in the tree has a key value x.key. In addition, we assign x.priority, which is a random number chosen independently for each node. We assume that all priorities are distinct and also that all keys are distinct. The nodes of the treap are ordered so that the keys obey the binarysearchtree property and the priorities obey the minheap order property:
 If v is a left child of u, then v.key < u.key.
 If v is a right child of u, then v.key > u.key.
 If v is a child of u, then v.priority > u.priority.
(This combination of properties is why the tree is called a “treap”: it has features of both a binary search tree and a heap.)
It helps to think of treaps in the following way. Suppose that we insert nodes \(x_1\), \(x_2\), …, \(x_n\), with associated keys, into a treap. Then the resulting treap is the tree that would have been formed if the nodes had been inserted into a normal binary search tree in the order given by their (randomly chosen) priorities, i.e., \(x_i\).priority < \(x_j\).priority means that we had inserted \(x_i\) before \(x_j\).
 a. Show that given a set of nodes \(x_1\), \(x_2\), …, \(x_n\), with associated keys and priorities, all distinct, the treap associated with these nodes is unique.
 b. Show that the expected height of a treap is Θ(lg n), and hence the expected time to search for a value in the treap is Θ(lg n).
Let us see how to insert a new node into an existing treap. The first thing we do is assign to the new node a random priority. Then we call the insertion algorithm, which we call TreapInsert, whose operation is illustrated in Figure 13.10.
 c. Explain how TreapInsert works. Explain the idea in English and give pseudocode. (Hint: Execute the usual binarysearchtree insertion procedure and then perform rotations to restore the minheap order property.)
 d. Show that the expected running time of TreapInsert is Θ(lg n).
TreapInsert performs a search and then a sequence of rotations. Although these two operations have the same expected running time, they have different costs in practice. A search reads information from the treap without modifying it. In contrast, a rotation changes parent and child pointers within the treap. On most computers, read operations are much faster than write operations. Thus we would like TreapInsert to perform few rotations. We will show that the expected number of rotations performed is bounded by a constant.
In order to do so, we will need some definitions, which Figure 13.11 depicts. The left spine of a binary search tree T is the simple path from the root to the node with the smallest key. In other words, the left spine is the simple path from the root that consists of only left edges. Symmetrically, the right spine of T is the simple path from the root consisting of only right edges. The length of a spine is the number of nodes it contains.
 e. Consider the treap T immediately after TreapInsert has inserted node x. Let C be the length of the right spine of the left subtree of x. Let D be the length of the left spine of the right subtree of x. Prove that the total number of rotations that were performed during the insertion of x is equal to C + D.
We will now calculate the expected values of C and D. Without loss of generality, we assume that the keys are 1, 2, …, n, since we are comparing them only to one another.
For nodes x and y in treap T, where y ≠ x, let k = x.key and i = y.key. We define indicator random variables
\(X_{i k}\) = I{y is in the right spine of the left subtree of x}.
f. Show that \(X_{i k}\) = 1 if and only if y.priority > x.priority, y.key < x.key, and, for every z such that y.key < z.key < x.key, we have y.priority < z.priority.
g. Show that
\(\begin{aligned} \Pr\left\lbrace X_{i k} = 1\right\rbrace &= \frac{(k  i  i)!}{(k  i + 1)!} \\ &= \frac{1}{(k  i + 1)(k  i)}. \end{aligned}\)
h. Show that
\(\begin{aligned} \operatorname{E}[C] &= ∑_{j = 1}^{k  1} \frac{1}{j (j + 1)} \\ &= 1  \frac{1}{k}. \end{aligned}\)
i. Use a symmetry argument to show that
E[D] = 1  \(\dfrac{1}{n  k + 1}\).
j. Conclude that the expected number of rotations performed when inserting a node into a treap is less than 2.

a. Proof by induction.
 Base case: If the treap has 0 node, the treap is trivially unique.
 Inductive case: If the treap has more than one nodes, the tree must have a root with the lowest priority, so the root is unique. The left subtree contains nodes that have keys less than key of root, by induction, we know the left subtree is unique. The same applies to the right subtree, so the treap is unique.

c. Solution is implemented here.
14 Augmenting Data Structures
14.1 Dynamic order statistics
14.11
Show how OSSelect(T.root, 10) operates on the redblack tree T of Figure 14.1.
Skipped.
14.12
Show how OSRank(T, x) operates on the redblack tree T of Figure 14.1 and the node x with x.key = 35.
Skipped.
14.13
Write a nonrecursive version of OSSelect.
OSSelect(x, i)
 r = x.left.size + 1
 while i ≠ r
 if i < r
 x = x.left
 else
 x = x.right
 i = i  r
 r = x.left.size + 1
 return x
14.14
Write a recursive procedure OSKeyRank(T, k) that takes as input an orderstatistic tree T and a key k and returns the rank of k in the dynamic set represented by T. Assume that the keys of T are distinct.
OSKeyRankHelper(x, k, a)
 if k == x.key
 return a + x.left.size + 1
 elseif k < T.key
 return OSKeyRankHelper(x.left, k, a)
 else
 return OSKeyRankHelper(x.right, k, a + x.left.size + 1)
OSKeyRank(T, k)
 return OSKeyRankHelper(T.root, k, 0)
14.15
Given an element x in an nnode orderstatistic tree and a natural number i, how can we determine the ith successor of x in the linear order of the tree in O(lg n) time?
OSIthSuccessor(x, i)
 if i == 0
 return x
 elseif x.right.size >= i
 return OSIthSuccessor(x.right, i  1)
 elseif x == x.p.left
 return OSIthSuccessor(x.p, i  x.right.size  1)
 else return OSIthSuccessor(x.p, i + x.left.size + 1)
14.16
Observe that whenever we reference the size attribute of a node in either OSSELECT or OSRank, we use it only to compute a rank. Accordingly, suppose we store in each node its rank in the subtree of which it is the root. Show how to maintain this information during insertion and deletion. (Remember that these two operations can cause rotations.)
Skipped.
14.17
Show how to use an orderstatistic tree to count the number of inversions (see Problem 24) in an array of size n in time O(n lg n).
Inversions(A)
 T = new empty orderstatistic redblack tree
 r = 0
 for i = 1 to A.length
 x = new node with key A[i]
 RBInsert(T, x)
 r = r + (i  OSRank(T, x))
 return r
14.18 ★
Consider n chords on a circle, each defined by its endpoints. Describe an O(n lg n) time algorithm to determine the number of pairs of chords that intersect inside the circle. (For example, if the n chords are all diameters that meet at the center, then the correct answer is \(\binom{n}{2}\) Assume that no two chords share an endpoint.
Define a chord as (a, b), where a is the endpoint of lesser degrees (the start endpoint), and b is the endpoint of greater degrees (the end endpoint), that is, a ∈ [0, 2 π), a ∈ [0, 2 π), and a ≤ b.
To get the number of pairs of chords that intersect inside the circle:

Sort these n chords by their start endpoint degree. This step can be done in O(n lg n) time.

Insert these sorted n chords into an orderstatistics redblack tree in order by their end endpoint degree.
Before a new chord (a, b) being inserted into the orderstatistics redblack tree, we know that all chords already in the orderstatistics redblack tree has start endpoints with lesser degree than a, because we have sorted the chords in the beginning. All chords that already in the tree that intersect with the new chord must have end endpoints within [a, b] range. Good thing that they are in an orderstatistics redblack tree with the end endpoint as key, so that we can find the number of chords with end endpoint within [a, b] range in O(lg k) time, where k is the number of nodes in the tree. The total running time of this step is O(n lg n).
So the total running time is in O(n lg n) limit.
14.2 How to augment a data structure
14.21
Show, by adding pointers to the nodes, how to support each of the dynamicset queries Minimum, Maximum, Successor, and Predecessor in O(1) worstcase time on an augmented orderstatistic tree. The asymptotic performance of other operations on orderstatistic trees should not be affected.
Skipped.
14.22
Can we maintain the blackheights of nodes in a redblack tree as attributes in the nodes of the tree without affecting the asymptotic performance of any of the redblack tree operations? Show how, or argue why not. How about maintaining the depths of nodes?
Blackheights of nodes can be maintained, but depth of nodes can not be maintained.
14.23 ★
Let ⊗ be an associative binary operator, and let a be an attribute maintained in each node of a redblack tree. Suppose that we want to include in each node x an additional attribute f such that x.f = \(x_1\).a ⊗ \(x_2\).a ⊗ ⋯ ⊗ \(x_m\).a, where \(x_1\), \(x_2\), …, \(x_m\) is the inorder listing of nodes in the subtree rooted at x. Show how to update the f attributes in O(1) time after a rotation. Modify your argument slightly to apply it to the size attributes in orderstatistic trees.
Assume we are doing right rotation:
┌───┐ ┌───┐
│ d │ │ b │
└─┬─┘ └─┬─┘
┌──┴───┐ ┌──┴───┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐
│ b │ │ e │ → │ a │ │ d │
└─┬─┘ └───┘ └───┘ └─┬─┘
┌──┴───┐ ┌──┴───┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐
│ a │ │ c │ │ c │ │ e │
└───┘ └───┘ └───┘ └───┘
Just follow this table:
Node  Old f attribute  New f attribute 

a  a.f  a.f 
b  a.f ⊗ b.a ⊗ c.f  a.f ⊗ b.a ⊗ c.f ⊗ d.a ⊗ e.f 
c  c.f  c.f 
d  a.f ⊗ b.a ⊗ c.f ⊗ d.a ⊗ e.f  c.f ⊗ d.a ⊗ e.f 
e  e.f  e.f 
14.24 ★
We wish to augment redblack trees with an operation RBEnumerate(x, a, b) that outputs all the keys k such that a ≤ k ≤ b in a redblack tree rooted at x. Describe how to implement RBEnumerate in Θ(m + lg n) time, where m is the number of keys that are output and n is the number of internal nodes in the tree. (Hint: You do not need to add new attributes to the redblack tree.)
RBEnumerate(x, a, b)
 if x ≠ nil
 if b < x.key
 RBEnumerate(x.left, a, b)
 elseif a > x.key
 RBEnumerate(x.right, a, b)
 else
 RBEnumerate(x.left, a, b)
 output x.key
 RBEnumerate(x.right, a, b)
14.3 Interval trees
14.31
Write pseudocode for LeftRotate that operates on nodes in an interval tree and updates the max attributes in O(1) time.
Skipped.
14.32
Rewrite the code for IntervalSearch so that it works properly when all intervals are open.
IntervalSearch(T, i)
 x = T.root
 while x ≠ T.nil and i does not overlap x.int
 if x.left ≠ T.nil and x.left.max > i.low
 x = x.left
 else x = x.right
 return x
14.33
Describe an efficient algorithm that, given an interval i, returns an interval overlapping i that has the minimum low endpoint, or T.nil if no such interval exists.
 x = T.root
 r = T.nil
 while x ≠ T.nil
 if i overlap x.int
 r = x
 if x.left ≠ T.nil and x.left.max ≥ i.low
 x = x.left
 else break
 else
 if x.left ≠ T.nil and x.left.max ≥ i.low
 x = x.left
 else x = x.right
 return r
14.34
Given an interval tree T and an interval i, describe how to list all intervals in T that overlap i in O(min(n, k lg n)) time, where k is the number of intervals in the output list. (Hint: One simple method makes several queries, modifying the tree between queries. A slightly more complicated method does not modify the tree.)
Skipped.
14.35
Suggest modifications to the intervaltree procedures to support the new operation IntervalSearchExactly(T, i) where T is an interval tree and i is an interval. The operation should return a pointer to a node x in T such that x.int.low = i.low and x.int.high = i.high, or T.nil if T contains no such node. All operations, including IntervalSearchExactly, should run in O(lg n) time on an nnode interval tree.
Skipped.
14.36
Show how to maintain a dynamic set Q of numbers that supports the operation MinGap, which gives the magnitude of the difference of the two closest numbers in Q. For example, if Q = {1, 5, 9, 15, 18, 22}, then MinGap(Q) returns 18  15 = 3, since 15 and 18 are the two closest numbers in Q. Make the operations Insert, Delete, Search, and MinGap as efficient as possible, and analyze their running times.
We add three extra augmenting property to every node:
 Minimal value of subtree min.
 Maximal value of subtree max.
 Minimal gap of numbers in the tree mingap.
For a node x:
 x.min =
 x.left.min if x.left ≠ nil
 x.key otherwise.
 x.max =
 x.right.max if x.right ≠ nil
 x.key otherwise.
 x.mingap = the minimal value of the following values if they exist:
 x.left.mingap
 x.right.mingap
 x.key  x.left.max
 x.right.min  x.key
 +∞
Then the operations are easy to implement.
14.37 ★
VLSI databases commonly represent an integrated circuit as a list of rectangles. Assume that each rectangle is rectilinearly oriented (sides parallel to the x and yaxes), so that we represent a rectangle by its minimum and maximum xand ycoordinates. Give an O(n lg n)time algorithm to decide whether or not a set of n rectangles so represented contains two rectangles that overlap. Your algorithm need not report all intersecting pairs, but it must report that an overlap exists if one rectangle entirely covers another, even if the boundary lines do not intersect. (Hint: Move a “sweep” line across the set of rectangles.)
As in the hint, we sweep one vertical line from left to right, and maintain a interval tree in which the intervals are the intervals that the sweep line intersects with the rectangles.
Before inserting each interval into the tree, check whether there is already interval in the tree that overlaps this new interval using IntervalSearch. If there exists any overlaps during inserting, then there are two rectangles that overlap.
14.X Problems
141 Point of maximum overlap
Suppose that we wish to keep track of a point of maximum overlap in a set of intervals—a point with the largest number of intervals in the set that overlap it.
a. Show that there will always be a point of maximum overlap that is an endpoint of one of the segments.
b. Design a data structure that efficiently supports the operations IntervalInsert, IntervalDelete, and FindPOM, which returns a point of maximum overlap. (Hint: Keep a redblack tree of all the endpoints. Associate a value of +1 with each left endpoint, and associate a value of 1 with each right endpoint. Augment each node of the tree with some extra information to maintain the point of maximum overlap.)

a. Skipped.

b. As in the hint, we got an redblack tree. We add an augmenting properties to each node:
 The sum of the all value property (the one with +1 and 1 values) rooted at that node sum.
 The point of maximum overlap pom.
 The number of overlaps on the point of maximum overlap maxoverlaps.
For a node x, we have:

x.sum = x.left.sum + x.value + x.right.sum

Depending on the maximal value of:
 x.left.maxoverlaps
 x.pom = x.left.pom
 x.maxoverlaps = x.left.maxoverlaps
 x.left.sum + x.value
 x.pom = x.key
 x.maxoverlaps = x.left.maxoverlaps + x.value
 x.left.sum + x.value + x.right.maxoverlaps
 x.pom = x.right.pom
 x.maxoverlaps = x.left.sum + x.value + x.right.maxoverlaps
 x.left.maxoverlaps
IntervalInsert, IntervalDelete are just normal augmented redblack tree operations. FindPOM just return the pom property of the root node.
142 Josephus permutation
We define the Josephus problem as follows. Suppose that n people form a circle and that we are given a positive integer m ≤ n. Beginning with a designated first person, we proceed around the circle, removing every mth person. After each person is removed, counting continues around the circle that remains. This process continues until we have removed all n people. The order in which the people are removed from the circle defines the (n, m)Josephus permutation of the integers 1, 2, …, n. For example, the (7, 3)Josephus permutation is ⟨3, 6, 2, 7, 5, 1, 4⟩.
a. Suppose that m is a constant. Describe an O(n)time algorithm that, given an integer n, outputs the (n, m)Josephus permutation.
b. Suppose that m is not a constant. Describe an O(n lg n)time algorithm that, given integers n and m, outputs the (n, m)Josephus permutation.
JosephusPermutation(n, m)
 T = new order statistic redblack tree with values 1, 2, …, n.
 i = 1
 for k = n downto 1
 i = (i + m  2) mod k + 1
 x = OSSelect(T.root, i)
 output x.key
 RBDelete(T, x)
IV Advanced Design and Analysis Techniques
15 Dynamic Programming
15.1 Rod cutting
15.11
Show that equation (15.4) follows from equation (15.3) and the initial condition T(0) = 1.
By induction:
Base case:
T(0)
= 1 + \(∑_{j = 0}^{1} T(j)\)
= 1
= \(2^0\).
Inductive case: If n > 0:
T(n)
= 1 + \(∑_{j = 0}^{n  1} T(j)\)
= 1 + \(∑_{j = 0}^{n  1} 2^j\) (By induction)
= 1 + (\(2^n\)  1)
= \(2^n\).
15.12
Show, by means of a counterexample, that the following “greedy” strategy does not always determine an optimal way to cut rods. Define the density of a rod of length i to be \(p_i\) / i, that is, its value per inch. The greedy strategy for a rod of length n cuts off a first piece of length i, where 1 ≤ i ≤ n, having maximum density. It then continues by applying the greedy strategy to the remaining piece of length n  i.
Assume we are cutting a rod of length 4, and we have:
 \(p_1\) = 1
 \(p_2\) = 6
 \(p_3\) = 10
 \(p_4\) = 12
Then the densities of length i is:
 \(d_1\) = 1
 \(d_2\) = 3
 \(d_3\) = 10 / 3
 \(d_4\) = 3
The greedy strategy give us 10 / 3 + 1 = 13 / 3 revenue, but if we cut the rod into two pieces of length 2, we can get revenue of 3 + 3 = 6.
15.13
Consider a modification of the rodcutting problem in which, in addition to a price \(p_i\) for each rod, each cut incurs a fixed cost of c. The revenue associated with a solution is now the sum of the prices of the pieces minus the costs of making the cuts. Give a dynamicprogramming algorithm to solve this modified problem.
Solution is implemented here.
15.14
Modify MemoizedCutRod to return not only the value but the actual solution, too.
Solution is implemented here.
15.15
The Fibonacci numbers are defined by recurrence (3.22). Give an O(n)time dynamicprogramming algorithm to compute the nth Fibonacci number. Draw the subproblem graph. How many vertices and edges are in the graph?
Solution is implemented here.
There are max(0, 2 n  2) edges.
15.2 Matrixchain multiplication
15.21
Find an optimal parenthesization of a matrixchain product whose sequence of dimensions is ⟨5, 10, 3, 12, 5, 50, 6⟩.
\(((A_1 A_2) ((A_3 A_4) (A_5 A_6)))\) (See here).
15.22
Give a recursive algorithm MatrixChainMultiply(A, s, i, j) that actually performs the optimal matrixchain multiplication, given the sequence of matrices ⟨\(A_1\), \(A_2\), …, \(A_n\)⟩, the s table computed by MatrixChainOrder, and the indices i and j. (The initial call would be MatrixChainMultiply(A, s, 1, n).)
Solution is implemented here.
15.23
Use the substitution method to show that the solution to the recurrence (15.6) is Ω(\(2^n\)).
Skipped.
15.24
Describe the subproblem graph for matrixchain multiplication with an input chain of length n. How many vertices does it have? How many edges does it have, and which edges are they?
The total number of vertices is:
\(∑_{i = 1}^n (n  i + 1)\)
= n (n + 1) / 2.
For subproblem of length k, there are 2 (k  1) subproblems, so the total number of edges is:
\(∑_{k = 1}^n ∑_{i = 1}^{n  k + 1} 2 (k  1)\)
= \(∑_{k = 1}^n 2 (n  k + 1) (k  1)\)
= (\(n^3\)  n) / 3.
15.25
Let R(i, j) be the number of times that table entry m[i, j] is referenced while computing other table entries in a call of MatrixChainOrder. Show that the total number of references for the entire table is
\(\displaystyle ∑_{i = 1}^n ∑_{j = i}^n R(i, j) = \frac{n^3  n}{3}\).
(Hint: You may find equation (A.3) useful.)
One edge in the subproblem graph means one reference, so the total number of references equals to the number of edges in the subproblem graph, which, according to exercise 15.24, is (\(n^3\)  n) / 3.
15.26
Show that a full parenthesization of an nelement expression has exactly n  1 pairs of parentheses.
Proof by induction.
Base case: A 1element expression has 0 pair of parentheses.
Inductive case: An nelement expression is composed of two subexpressions. Suppose the left one has k elements, then the right one has n  k elements. By induction, we know these two subexpressions have k  1 and n  k  1 pairs of parentheses. Since nelement expression is enclosed inside a pair of parentheses, the total pairs of parentheses is (k  1) + (n  k  1) + 1 = n  1.
15.3 Elements of dynamic programming
15.31
Which is a more efficient way to determine the optimal number of multiplications in a matrixchain multiplication problem: enumerating all the ways of parenthesizing the product and computing the number of multiplications for each, or running RecursiveMatrixChain? Justify your answer.
Skipped.
15.32
Draw the recursion tree for the MergeSort procedure from Section 2.3.1 on an array of 16 elements. Explain why memoization fails to speed up a good divideandconquer algorithm such as MergeSort.
Because the subproblems do not overlap.
15.33
Consider a variant of the matrixchain multiplication problem in which the goal is to parenthesize the sequence of matrices so as to maximize, rather than minimize, the number of scalar multiplications. Does this problem exhibit optimal substructure?
Yes.
15.34
As stated, in dynamic programming we first solve the subproblems and then choose which of them to use in an optimal solution to the problem. Professor Capulet claims that we do not always need to solve all the subproblems in order to find an optimal solution. She suggests that we can find an optimal solution to the matrixchain multiplication problem by always choosing the matrix \(A_k\) at which to split the subproduct \(A_i A_{i + 1} ⋯ A_j\) (by selecting k to minimize the quantity \(p_{i  1} p_k p_j\)) before solving the subproblems. Find an instance of the matrixchain multiplication problem for which this greedy approach yields a suboptimal solution.
Skipped.
15.35
Suppose that in the rodcutting problem of Section 15.1, we also had limit \(l_i\) on the number of pieces of length i that we are allowed to produce, for i = 1, 2, …, n. Show that the optimalsubstructure property described in Section 15.1 no longer holds.
For rod of length n, the number of ways to cutting it depends on the number of pieces has already been cut off, not the only the value of n.
15.36
Imagine that you wish to exchange one currency for another. You realize that instead of directly exchanging one currency for another, you might be better off making a series of trades through other currencies, winding up with the currency you want. Suppose that you can trade n different currencies, numbered 1, 2, …, n, where you start with currency 1 and wish to wind up with currency n. You are given, for each pair of currencies i and j, an exchange rate \(r_{i j}\), meaning that if you start with d units of currency i, you can trade for \(d r_{i j}\) units of currency j. A sequence of trades may entail a commission, which depends on the number of trades you make. Let \(c_k\) be the commission that you are charged when you make k trades. Show that, if \(c_k\) = 0 for all k = 1, 2, …, n, then the problem of finding the best sequence of exchanges from currency 1 to currency n exhibits optimal substructure. Then show that if commissions \(c_k\) are arbitrary values, then the problem of finding the best sequence of exchanges from currency 1 to currency n does not necessarily exhibit optimal substructure.
Skipped.
15.4 Longest common subsequence
15.41
Determine an LCS of ⟨1, 0, 0, 1, 0, 1, 0, 1⟩ and ⟨0, 1, 0, 1, 1, 0, 1, 1, 0⟩.
The result is ⟨1, 0, 0, 1, 1, 0⟩.
See here.
15.42
Give pseudocode to reconstruct an LCS from the completed c table and the original sequences X = ⟨\(x_1\), \(x_2\), …, \(x_m\)⟩ and Y = ⟨\(y_1\), \(y_2\), …, \(y_n\)⟩ in O(m + n) time, without using the b table.
Solution is implemented here.
15.43
Give a memoized version of LCSLength that runs in O(m n) time.
Solution is implemented here.
15.44
Show how to compute the length of an LCS using only 2 ⋅ min(m, n) entries in the c table plus O(1) additional space. Then show how to do the same thing, but using min(m, n) entries plus O(1) additional space.
Solution is implemented here.
15.45
Give an O(\(n^2\))time algorithm to find the longest monotonically increasing subsequence of a sequence of n numbers.
Solution is implemented here.
15.46 ★
Give an O(n lg n)time algorithm to find the longest monotonically increasing subsequence of a sequence of n numbers. (Hint: Observe that the last element of a candidate subsequence of length i is at least as large as the last element of a candidate subsequence of length i  1. Maintain candidate subsequences by linking them through the input sequence.)
Solution is implemented here.
15.5 Optimal binary search trees
15.51
Write pseudocode for the procedure ConstructOptimalBST(root) which, given the table root, outputs the structure of an optimal binary search tree. For the example in Figure 15.10, your procedure should print out the structure
 \(k_2\) is the root
\(k_1\) is the left child of \(k_2\)
\(d_0\) is the left child of \(k_1\)
\(d_1\) is the right child of \(k_1\)
\(k_5\) is the right child of \(k_2\)
\(k_4\) is the left child of \(k_5\)
\(k_3\) is the left child of \(k_4\)
\(d_2\) is the left child of \(k_3\)
\(d_3\) is the right child of \(k_3\)
\(d_4\) is the right child of \(k_4\)
\(d_5\) is the right child of \(k_5\)corresponding to the optimal binary search tree shown in Figure 15.9(b).
Solution is implemented here.
15.52
Determine the cost and structure of an optimal binary search tree for a set of n = 7 keys with the following probabilities:
i 0 1 2 3 4 5 6 7 \(p_i\) 0.04 0.06 0.08 0.02 0.10 0.12 0.14 \(q_i\) 0.06 0.06 0.06 0.06 0.05 0.05 0.05 0.05
The optimal cost is 3.12. The optimal structure is:
┌─────┐
│ k_5 │
└──┬──┘
┌──────────────┴───────────────┐
┌──┴──┐ ┌──┴──┐
│ k_2 │ │ k_7 │
└──┬──┘ └──┬──┘
┌────────┴────────┐ ┌───┴────┐
┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐
│ k_1 │ │ k_3 │ │ k_6 │ │ d_7 │
└──┬──┘ └──┬──┘ └──┬──┘ └─────┘
┌───┴────┐ ┌───┴────┐ ┌───┴────┐
┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐
│ d_0 │ │ d_1 │ │ d_2 │ │ k_4 │ │ d_5 │ │ d_6 │
└─────┘ └─────┘ └─────┘ └──┬──┘ └─────┘ └─────┘
┌───┴────┐
┌──┴──┐ ┌──┴──┐
│ d_3 │ │ d_4 │
└─────┘ └─────┘
See here.
15.53
Suppose that instead of maintaining the table w[i, j], we computed the value of w(i, j) directly from equation (15.12) in line 9 of OptimalBST and used this computed value in line 11. How would this change affect the asymptotic running time of OptimalBST?
This change doesn’t affect the asymptotic running time of OptimalBST. because computing w(i, j) takes the same time as the innermost loop.
15.54 ★
Knuth [212] has shown that there are always roots of optimal subtrees such that root[i, j  1] ≤ root[i, j] ≤ root[i + 1, j] for all 1 ≤ i < j ≤ n. Use this fact to modify the OptimalBST procedure to run in Θ(\(n^2\)) time.
Solution is implemented here.
15.X Problems
151 Longest simple path in a directed acyclic graph
Suppose that we are given a directed acyclic graph G = (V, E) with realvalued edge weights and two distinguished vertices s and t. Describe a dynamic programming approach for finding a longest weighted simple path from s to t. What does the subproblem graph look like? What is the efficiency of your algorithm?
Solution is implemented here.
The running time is O(V + E).
152 Longest palindrome subsequence
A palindrome is a nonempty string over some alphabet that reads the same forward and backward. Examples of palindromes are all strings of length 1,
civic
,racecar
, andaibohphobia
(fear of palindromes).Give an efficient algorithm to find the longest palindrome that is a subsequence of a given input string. For example, given the input
character
, your algorithm should returncarac
. What is the running time of your algorithm?
Solution is implemented here.
The running time is O(\(n^2\)), where n is the length of the input.
153 Bitonic euclidean travelingsalesman problem
In the euclidean travelingsalesman problem, we are given a set of n points in the plane, and we wish to find the shortest closed tour that connects all n points. Figure 15.11(a) shows the solution to a 7point problem. The general problem is NPhard, and its solution is therefore believed to require more than polynomial time (see Chapter 34).
J. L. Bentley has suggested that we simplify the problem by restricting our attention to bitonic tours, that is, tours that start at the leftmost point, go strictly rightward to the rightmost point, and then go strictly leftward back to the starting point. Figure 15.11(b) shows the shortest bitonic tour of the same 7 points. In this case, a polynomialtime algorithm is possible.
Describe an O(\(n^2\))time algorithm for determining an optimal bitonic tour. You may assume that no two points have the same xcoordinate and that all operations on real numbers take unit time. (Hint: Scan left to right, maintaining optimal possibilities for the two parts of the tour.)
Solution is implemented here.
154 Printing neatly
Consider the problem of neatly printing a paragraph with a monospaced font (all characters having the same width) on a printer. The input text is a sequence of n words of lengths \(l_1\), \(l_2\), …, \(l_n\), measured in characters. We want to print this paragraph neatly on a number of lines that hold a maximum of M characters each. Our criterion of “neatness” is as follows. If a given line contains words i through j, where i ≤ j, and we leave exactly one space between words, the number of extra space characters at the end of the line is M  j + i  \(∑_{k = i}^j l_k\), which must be nonnegative so that the words fit on the line. We wish to minimize the sum, over all lines except the last, of the cubes of the numbers of extra space characters at the ends of lines. Give a dynamicprogramming algorithm to print a paragraph of n words neatly on a printer. Analyze the running time and space requirements of your algorithm.
Solution is implemented here.
Running time is O(M n), space requirement is O(n).
155 Edit distance
In order to transform one source string of text x[1‥m] to a target string y[1‥n], we can perform various transformation operations. Our goal is, given x and y, to produce a series of transformations that change x to y. We use an array z—assumed to be large enough to hold all the characters it will need—to hold the intermediate results. Initially, z is empty, and at termination, we should have z[j] = y[j] for j = 1, 2, …, n. We maintain current indices i into x and j into z, and the operations are allowed to alter z and these indices. Initially, i = j = 1. We are required to examine every character in x during the transformation, which means that at the end of the sequence of transformation operations, we must have i = m + 1.
We may choose from among six transformation operations:
 Copy a character from x to z by setting z[j] = x[i] and then incrementing both i and j. This operation examines x[i].
 Replace a character from x by another character c, by setting z[j] = c, and then incrementing both i and j. This operation examines x[i].
 Delete a character from x by incrementing i but leaving j alone. This operation examines x[i].
 Insert the character c into z by setting z[j] = c and then incrementing j, but leaving i alone. This operation examines no characters of x.
 Twiddle (i.e., exchange) the next two characters by copying them from x to z but in the opposite order; we do so by setting z[j] = x[i + 1] and z[j + 1] = x[i] and then setting i = i + 2 and j = j + 2. This operation examines x[i] and x[i + 1].
 Kill the remainder of x by setting i = m + 1. This operation examines all characters in x that have not yet been examined. This operation, if performed, must be the final operation.
As an example, one way to transform the source string
algorithm
to the target stringaltruistic
is to use the following sequence of operations, where the underlined characters are x[i] and z[j] after the operation:
Operation x z initial strings a̲lgorithm
_
copy al̲gorithm
a_
copy alg̲orithm
al_
replace by t
algo̲rithm
alt_
delete algor̲ithm
alt_
copy algori̲thm
altr_
insert u
algori̲thm
altru_
insert i
algori̲thm
altrui_
insert s
algori̲thm
altruis_
twiddle algorith̲m
altruisti_
insert c
algorith̲m
altruistic_
kill algorithm_
altruistic_
Note that there are several other sequences of transformation operations that transform
algorithm
toaltruistic
.Each of the transformation operations has an associated cost. The cost of an operation depends on the specific application, but we assume that each operation’s cost is a constant that is known to us. We also assume that the individual costs of the copy and replace operations are less than the combined costs of the delete and insert operations; otherwise, the copy and replace operations would not be used. The cost of a given sequence of transformation operations is the sum of the costs of the individual operations in the sequence. For the sequence above, the cost of transforming
algorithm
toaltruistic
is(3 ⋅ cost(copy)) + cost(replace) + cost(delete) + (4 ⋅ cost(insert)) + cost(twiddle) + cost(kill).
 a. Given two sequences x[1‥m] and y[1‥n] and set of transformationoperation costs, the edit distance from x to y is the cost of the least expensive operation sequence that transforms x to y. Describe a dynamicprogramming algorithm that finds the edit distance from x[1‥m] to y[1‥n] and prints an optimal operation sequence. Analyze the running time and space requirements of your algorithm.
The editdistance problem generalizes the problem of aligning two DNA sequences (see, for example, Setubal and Meidanis [310, Section 3.2]). There are several methods for measuring the similarity of two DNA sequences by aligning them. One such method to align two sequences x and y consists of inserting spaces at arbitrary locations in the two sequences (including at either end) so that the resulting sequences x′ and y′ have the same length but do not have a space in the same position (i.e., for no position j are both x′[j] and y′[j] a space). Then we assign a “score” to each position. Position j receives a score as follows:
 +1 if x′[j] = y′[j] and neither is a space,
 1 if x′[j] ≠ y′[j] and neither is a space,
 2 if either x′[j] or y′[j] is a space.
The score for the alignment is the sum of the scores of the individual positions. For example, given the sequences x =
GATCGGCAT
and y =CAATGTGAATC
, one alignment isG ATCG GCAT CAAT GTGAATC *++*+*+++*
A
+
under a position indicates a score of +1 for that position, a
indicates a score of 1, and a*
indicates a score of 2, so that this alignment has a total score of 6 ⋅ 1  2 ⋅ 1  4 ⋅ 2 = 4.
 b. Explain how to cast the problem of finding an optimal alignment as an edit distance problem using a subset of the transformation operations copy, replace, delete, insert, twiddle, and kill.
Solution is implemented here.

a. Running time is in Θ(m n), space requirement is Θ(m n).

b. Each alignment can be transformed into a sequence of transformation operations:
 The +1 case can be transformed into copy operation with cost of 1.
 The 1 case can be transformed into replace operation with cost of 1.
 The 2 case can be transformed into insert or delete operation with cost of 2.
And we set twiddle and kill operation as cost of ∞ since there is no corresponding alignment.
Note that the cost is the negative score, since our algorithm of editdistance minimizes the cost but we want to maximize the score.
156 Planning a company party
Professor Stewart is consulting for the president of a corporation that is planning a company party. The company has a hierarchical structure; that is, the supervisor relation forms a tree rooted at the president. The personnel office has ranked each employee with a conviviality rating, which is a real number. In order to make the party fun for all attendees, the president does not want both an employee and his or her immediate supervisor to attend.
Professor Stewart is given the tree that describes the structure of the corporation, using the leftchild, rightsibling representation described in Section 10.4. Each node of the tree holds, in addition to the pointers, the name of an employee and that employee’s conviviality ranking. Describe an algorithm to make up a guest list that maximizes the sum of the conviviality ratings of the guests. Analyze the running time of your algorithm.
Solution is implemented here.
Running time is Θ(n) where n is the total number of employees.
157 Viterbi algorithm
We can use dynamic programming on a directed graph G = (V, E) for speech recognition. Each edge (u, v) ∈ E is labeled with a sound σ(u, v) from a finite set Σ of sounds. The labeled graph is a formal model of a person speaking a restricted language. Each path in the graph starting from a distinguished vertex \(v_0\) ∈ V corresponds to a possible sequence of sounds produced by the model. We define the label of a directed path to be the concatenation of the labels of the edges on that path.
 a. Describe an efficient algorithm that, given an edgelabeled graph G with distinguished vertex \(v_0\) and a sequence s = ⟨\(\sigma_1\), \(\sigma_2\), …, \(\sigma_k\)⟩ of sounds from Σ, returns a path in G that begins at \(v_0\) and has s as its label, if any such path exists. Otherwise, the algorithm should return nosuchpath. Analyze the running time of your algorithm. (Hint: You may find concepts from Chapter 22 useful.)
Now, suppose that every edge (u, v) ∈ E has an associated nonnegative probability p(u, v) of traversing the edge (u, v) from vertex u and thus producing the corresponding sound. The sum of the probabilities of the edges leaving any vertex equals 1. The probability of a path is defined to be the product of the probabilities of its edges. We can view the probability of a path beginning at \(v_0\) as the probability that a “random walk” beginning at \(v_0\) will follow the specified path, where we randomly choose which edge to take leaving a vertex u according to the probabilities of the available edges leaving u.
 b. Extend your answer to part (a) so that if a path is returned, it is a most probable path starting at \(v_0\) and having label s. Analyze the running time of your algorithm.
Solution is implemented here.
158 Image compression by seam carving
We are given a color picture consisting of an m × n array A[1‥m, 1‥n] of pixels, where each pixel specifies a triple of red, green, and blue (RGB) intensities. Suppose that we wish to compress this picture slightly. Specifically, we wish to remove one pixel from each of the m rows, so that the whole picture becomes one pixel narrower. To avoid disturbing visual effects, however, we require that the pixels removed in two adjacent rows be in the same or adjacent columns; the pixels removed form a “seam” from the top row to the bottom row where successive pixels in the seam are adjacent vertically or diagonally.
a. Show that the number of such possible seams grows at least exponentially in m, assuming that n > 1.
b. Suppose now that along with each pixel A[i, j], we have calculated a realvalued disruption measure d[i, j], indicating how disruptive it would be to remove pixel A[i, j]. Intuitively, the lower a pixel’s disruption measure, the more similar the pixel is to its neighbors. Suppose further that we define the disruption measure of a seam to be the sum of the disruption measures of its pixels.
Give an algorithm to find a seam with the lowest disruption measure. How efficient is your algorithm?

a. Assume we are selecting a seam from top to bottom. First, we can select one of n pixels in the first row, then for each pixel we already selected, there are two or three pixels in the next row to select from. So the total number of seams is at least n × \(2^{m  1}\) = n / 2 × \(2^m\), which is at least exponentially in m.

b. Solution is implemented here.
159 Breaking a string
A certain stringprocessing language allows a programmer to break a string into two pieces. Because this operation copies the string, it costs n time units to break a string of n characters into two pieces. Suppose a programmer wants to break a string into many pieces. The order in which the breaks occur can affect the total amount of time used. For example, suppose that the programmer wants to break a 20character string after characters 2, 8, and 10 (numbering the characters in ascending order from the lefthand end, starting from 1). If she programs the breaks to occur in lefttoright order, then the first break costs 20 time units, the second break costs 18 time units (breaking the string from characters 3 to 20 at character 8), and the third break costs 12 time units, totaling 50 time units. If she programs the breaks to occur in righttoleft order, however, then the first break costs 20 time units, the second break costs 10 time units, and the third break costs 8 time units, totaling 38 time units. In yet another order, she could break first at 8 (costing 20), then break the left piece at 2 (costing 8), and finally the right piece at 10 (costing 12), for a total cost of 40.
Design an algorithm that, given the numbers of characters after which to break, determines a leastcost way to sequence those breaks. More formally, given a string S with n characters and an array L[1‥m] containing the break points, compute the lowest cost for a sequence of breaks, along with a sequence of breaks that achieves this cost.
Solution is implemented here.
1510 Planning an investment strategy
Your knowledge of algorithms helps you obtain an exciting job with the Acme Computer Company, along with a $10,000 signing bonus. You decide to invest this money with the goal of maximizing your return at the end of 10 years. You decide to use the Amalgamated Investment Company to manage your investments. Amalgamated Investments requires you to observe the following rules. It offers n different investments, numbered 1 through n. In each year j, investment i provides a return rate of \(r_{i j}\). In other words, if you invest d dollars in investment i in year j, then at the end of year j, you have \(d r_{i j}\) dollars. The return rates are guaranteed, that is, you are given all the return rates for the next 10 years for each investment. You make investment decisions only once per year. At the end of each year, you can leave the money made in the previous year in the same investments, or you can shift money to other investments, by either shifting money between existing investments or moving money to a new investment. If you do not move your money between two consecutive years, you pay a fee of \(f_1\) dollars, whereas if you switch your money, you pay a fee of \(f_2\) dollars, where \(f_2\) > \(f_1\).
 a. The problem, as stated, allows you to invest your money in multiple investments in each year. Prove that there exists an optimal investment strategy that, in each year, puts all the money into a single investment. (Recall that an optimal investment strategy maximizes the amount of money after 10 years and is not concerned with any other objectives, such as minimizing risk.)
 b. Prove that the problem of planning your optimal investment strategy exhibits optimal substructure.
 c. Design an algorithm that plans your optimal investment strategy. What is the running time of your algorithm?
 d. Suppose that Amalgamated Investments imposed the additional restriction that, at any point, you can have no more than $15,000 in any one investment. Show that the problem of maximizing your income at the end of 10 years no longer exhibits optimal substructure.
 a. Skipped.
 b. Skipped.
 c. Solution is implemented here. Running time is Θ(m n), where m is the number of total years, which equals to 10 in this problem.
 d. Skipped.
1511 Inventory planning
The Rinky Dink Company makes machines that resurface ice rinks. The demand for such products varies from month to month, and so the company needs to develop a strategy to plan its manufacturing given the fluctuating, but predictable, demand. The company wishes to design a plan for the next n months. For each month i, the company knows the demand \(d_i\), that is, the number of machines that it will sell. Let D = \(∑_{i = 1}^n d_i\) be the total demand over the next n months. The company keeps a fulltime staff who provide labor to manufacture up to m machines per month. If the company needs to make more than m machines in a given month, it can hire additional, parttime labor, at a cost that works out to c dollars per machine. Furthermore, if, at the end of a month, the company is holding any unsold machines, it must pay inventory costs. The cost for holding j machines is given as a function h(j) for j = 1, 2, …, D, where h(j) ≥ 0 for 1 ≤ j ≤ D and h(j) ≤ h(j + 1) for 1 ≤ j ≤ D  1.
Give an algorithm that calculates a plan for the company that minimizes its costs while fulfilling all the demand. The running time should be polynomial in n and D.
Solution is implemented here.
1512 Signing freeagent baseball players
Suppose that you are the general manager for a majorleague baseball team. During the offseason, you need to sign some freeagent players for your team. The team owner has given you a budget of $X to spend on free agents. You are allowed to spend less than $X altogether, but the owner will fire you if you spend any more than $X.
You are considering N different positions, and for each position, P freeagent players who play that position are available. Because you do not want to overload your roster with too many players at any position, for each position you may sign at most one free agent who plays that position. (If you do not sign any players at a particular position, then you plan to stick with the players you already have at that position.)
To determine how valuable a player is going to be, you decide to use a sabermetric statistic known as “VORP,” or “value over replacement player.” A player with a higher VORP is more valuable than a player with a lower VORP. A player with a higher VORP is not necessarily more expensive to sign than a player with a lower VORP, because factors other than a player’s value determine how much it costs to sign him.
For each available freeagent player, you have three pieces of information:
 the player’s position,
 the amount of money it will cost to sign the player, and
 the player’s VORP.
Devise an algorithm that maximizes the total VORP of the players you sign while spending no more than $X altogether. You may assume that each player signs for a multiple of $100,000. Your algorithm should output the total VORP of the players you sign, the total amount of money you spend, and a list of which players you sign. Analyze the running time and space requirement of your algorithm.
Solution is implemented here.
16 Greedy Algorithms
16.1 An activityselection problem
16.11
Give a dynamicprogramming algorithm for the activityselection problem, based on recurrence (16.2). Have your algorithm compute the sizes c[i, j] as defined above and also produce the maximumsize subset of mutually compatible activities. Assume that the inputs have been sorted as in equation (16.1). Compare the running time of your solution to the running time of GreedyActivitySelector.
Skipped.
16.12
Suppose that instead of always selecting the first activity to finish, we instead select the last activity to start that is compatible with all previously selected activities. Describe how this approach is a greedy algorithm, and prove that it yields an optimal solution.
This is essentially the same as the original problem, but with time reversed.
16.13
Not just any greedy approach to the activityselection problem produces a maximumsize set of mutually compatible activities. Give an example to show that the approach of selecting the activity of least duration from among those that are compatible with previously selected activities does not work. Do the same for the approaches of always selecting the compatible activity that overlaps the fewest other remaining activities and always selecting the compatible remaining activity with the earliest start time.

Least duration:
▬▬▬▬ ▬▬▬▬ ▬▬▬

Fewest overlaps:
▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬ ▬▬▬

Earliest start time:
▬▬▬▬▬▬▬▬▬ ▬▬▬ ▬▬▬
16.14
Suppose that we have a set of activities to schedule among a large number of lecture halls, where any activity can take place in any lecture hall. We wish to schedule all the activities using as few lecture halls as possible. Give an efficient greedy algorithm to determine which activity should use which lecture hall.
(This problem is also known as the intervalgraph coloring problem. We can create an interval graph whose vertices are the given activities and whose edges connect incompatible activities. The smallest number of colors required to color every vertex so that no two adjacent vertices have the same color corresponds to finding the fewest lecture halls needed to schedule all of the given activities.)
Solution is implemented here.
16.15
Consider a modification to the activityselection problem in which each activity \(a_i\) has, in addition to a start and finish time, a value \(v_i\). The objective is no longer to maximize the number of activities scheduled, but instead to maximize the total value of the activities scheduled. That is, we wish to choose a set A of compatible activities such that \(Σ_{a_k ∈ A} v_k\) is maximized. Give a polynomialtime algorithm for this problem.
Solution is implemented here.
16.2 Elements of the greedy strategy
16.21
Prove that the fractional knapsack problem has the greedychoice property.
For each subproblem, as long as there are items that with more value per pound than some items already in the knapsack, we can replace some items already in the knapsack with items with more value per pound to get a better solution. So for each subproblem, we must put as much as possible items with the most value per pound.
16.22
Give a dynamicprogramming solution to the 01 knapsack problem that runs in O(n W) time, where n is the number of items and W is the maximum weight of items that the thief can put in his knapsack.
Solution is implemented here.
16.23
Suppose that in a 01 knapsack problem, the order of the items when sorted by increasing weight is the same as their order when sorted by decreasing value. Give an efficient algorithm to find an optimal solution to this variant of the knapsack problem, and argue that your algorithm is correct.
Algorithm: For each subproblem, select the item with most value within the remaining items.
Proof: Like 16.21, if we have selected other item instead of the item with most value, we can replace that item with the item with most value to get more optimized solution.
16.24
Professor Gekko has always dreamed of inline skating across North Dakota. He plans to cross the state on highway U.S. 2, which runs from Grand Forks, on the eastern border with Minnesota, to Williston, near the western border with Montana. The professor can carry two liters of water, and he can skate m miles before running out of water. (Because North Dakota is relatively flat, the professor does not have to worry about drinking water at a greater rate on uphill sections than on flat or downhill sections.) The professor will start in Grand Forks with two full liters of water. His official North Dakota state map shows all the places along U.S. 2 at which he can refill his water and the distances between these locations.
The professor’s goal is to minimize the number of water stops along his route across the state. Give an efficient method by which he can determine which water stops he should make. Prove that your strategy yields an optimal solution, and give its running time.
Solution is implemented here.
The strategy is to stop as late as possible at each step. Suppose the latest water stop is k. If you stop early, then when you reach k, you can choose to stop or continue.
 If you stop, you will have one more stop than stopping at k directly.
 If you continue, you will have the same number of stops so far, but with less water. Less water means earlier stop, so you can go no further than stopping at k directly.
So the only choice is to stop at k.
The running time is O(n).
16.25
Describe an efficient algorithm that, given a set {\(x_1\), \(x_2\), …, \(x_n\)} of points on the real line, determines the smallest set of unitlength closed intervals that contains all of the given points. Argue that your algorithm is correct.
Solution is implemented here.
16.26 ★
Show how to solve the fractional knapsack problem in O(n) time.
Solution is implemented here.
16.27
Suppose you are given two sets A and B, each containing n positive integers. You can choose to reorder each set however you like. After reordering, let \(a_i\) be the ith element of set A, and let \(b_i\) be the ith element of set B. You then receive a payoff of \(∏_{i = 1}^n a_i^{b_i}\). Give an algorithm that will maximize your payoff. Prove that your algorithm maximizes the payoff, and state its running time.
Solution is implemented here.
Just sort two sets incrementally.
Proof: Let \(a_i\) and \(a_j\) be two elements from set A, and \(a_i\) ≤ \(a_j\); Let \(b_i\) and \(b_j\) be two elements from set B, and \(b_i\) ≤ \(b_j\), we have:
\(\dfrac{a_i^{b_i} a_j^{b_j}}{a_i^{b_j} a_j^{b_i}}\)
= \(a_i^{b_i  b_j} a_j^{b_j  b_i}\)
= \(\dfrac{a_j^{b_j  b_i}}{a_i^{b_j  b_i}}\)
= \(\left(\dfrac{a_j}{a_i}\right)^{b_j  b_i}\).
Since \(a_i\), \(a_j\), \(b_i\) and \(b_j\) are positive integers, we have \(\frac{a_j}{a_i}\) ≥ 1, and \(b_j  b_i\) ≥ 0, then:
\(\left(\dfrac{a_j}{a_i}\right)^{b_j  b_i}\) ≥ 1, which means \(a_i^{b_i} a_j^{b_j}\) ≥ \(a_i^{b_j} a_j^{b_i}\).
So for each pair \(a_i\) and \(a_j\) in set A, if \(a_i\) ≤ \(a_j\), we must ensure that the corresponding \(b_i\) ≤ \(b_j\). To get the maximize payoff, we should make sure that the relative order in two sets are the same. The simplest way to do this is sorting two sets in increasing order.
16.31
Explain why, in the proof of Lemma 16.2, if x.freq = b.freq, then we must have a.freq = b.freq = x.freq = y.freq.
Since b.freq = x.freq ≤ y.freq, and y.freq ≤ b.freq, we have x.freq = y.freq = b.freq.
Since b.freq = x.freq ≤ a.freq, and a.freq ≤ b.freq, we have a.freq = b.freq = x.freq.
So a.freq = b.freq = x.freq = y.freq.
16.32
Prove that a binary tree that is not full cannot correspond to an optimal prefix code.
Suppose there is a node x with only one child y, then we can use the codeword for x to represent y directly, since we know exactly what the last bit will be, we don’t need to include it explicitly. So we can transform the original tree into a new tree with one leaf node’s depth shortened by 1. So the original tree cannot correspond to an optimal prefix code.
16.33
What is an optimal Huffman code for the following set of frequencies, based on the first 8 Fibonacci numbers?
a:1 b:1 c:2 d:3 e:5 f:8 g:13 h:21
Can you generalize your answer to find the optimal code when the frequencies are the first n Fibonacci numbers?
Character  Huffman code 

a 
0000000 
b 
0000001 
c 
000001 
d 
00001 
e 
0001 
f 
001 
g 
01 
h 
1 
If there are n Fibonacci frequencies, then:
 The Huffman code for the first frequency is (n  1)
0
s.  The Huffman code for the ith (i > 1) frequency is (n  i)
0
s, followed by a1
.
16.34
Prove that we can also express the total cost of a tree for a code as the sum, over all internal nodes, of the combined frequencies of the two children of the node.
Proof by induction:

Base case: For empty nodes (nil), there are no leaf nodes or internal nodes, so the cost and the sum of the frequencies of internal nodes are both 0, the claim holds. We can assume the frequency of a nil node is 0.

Inductive case: For nonnil nodes:
 Let the node be T;
 Let the left child be L;
 Let the right child be R;
 Let the total frequency of leaf nodes of the left child be f;
 Let the total frequency of leaf nodes of the right child be g;
 Let the total sum of frequencies of internal nodes of left child be s;
 Let the total sum of frequencies of internal nodes of right child be t.
The cost of the node is:
\(∑_{c ∈ T} c{.}\textit{freq} ⋅ d_T(c)\)
= \(∑_{c ∈ L} c{.}\textit{freq} ⋅ d_T(c) + ∑_{c ∈ R} c{.}\textit{freq} ⋅ d_T(c)\)
= \(∑_{c ∈ L} c{.}\textit{freq} ⋅ (d_L(c) + 1) + ∑_{c ∈ R} c{.}\textit{freq} ⋅ (d_R(c) + 1)\)
= \(∑_{c ∈ L} c{.}\textit{freq} ⋅ d_L(c) + ∑_{c ∈ R} c{.}\textit{freq} ⋅ d_R(c) + ∑_{c ∈ L} c{.}\textit{freq} + ∑_{c ∈ R} c{.}\textit{freq}\)
= s + t + f + g (by induction)Since the frequency of T = f + g, so s + t + f + g equals to the sum of frequencies of internal nodes of T. The claim holds.
16.35
Prove that if we order the characters in an alphabet so that their frequencies are monotonically decreasing, then there exists an optimal code whose codeword lengths are monotonically increasing.
Proof by contradiction:
Suppose all optimal codes are nonmonotonically increasing, then there must exist a pair of codes in which the longer coder has lesser frequency.
 Let s be the length of the shorter code;
 Let t be the length of the longer code;
 Let f be the frequency of the shorter code;
 Let g be the frequency of the longer code.
We have s < t and f > g.
We can represent the cost as c + s f + t g, where c is some constant value.
Now we devise a new code which swaps the representation code of the selected pair. Now we get a new code whose cost is c + s g + t f.
(c + s f + t g)  (c + s g + t f)
= (c + s f + t g)  (c + s g + t f)
= s (f  g) + t (f  g)
= (s + t) (f  g)
> 0.
So we got a code which is lesser than the original code, so the original code is not optimal. So our supposition is wrong, which means there exists an optimal code whose codeword lengths are monotonically increasing.
16.36
Suppose we have an optimal prefix code on a set C = {0, 1, …, n  1} of characters and we wish to transmit this code using as few bits as possible. Show how to represent any optimal prefix code on C using only 2 n  1 + n ⌈lg n⌉ bits. (Hint: Use 2 n  1 bits to specify the structure of the tree, as discovered by a walk of the tree.)
Solution is implemented here.
16.37
Generalize Huffman’s algorithm to ternary codewords (i.e., codewords using the symbols 0, 1, and 2), and prove that it yields optimal ternary codes.
Skipped.
16.38
Suppose that a data file contains a sequence of 8bit characters such that all 256 characters are about equally common: the maximum character frequency is less than twice the minimum character frequency. Prove that Huffman coding in this case is no more efficient than using an ordinary 8bit fixedlength code.
Skipped.
16.39
Show that no compression scheme can expect to compress a file of randomly chosen 8bit characters by even a single bit. (Hint: Compare the number of possible files with the number of possible encoded files.)
Assume there exists such a compression scheme.
The number of possible files with n bits is \(2^n\), but the number of possible files with n  1 bits is \(2^{n  1}\), which is less than \(2^n\), which means that some nbits files will be transformed into the same (n  1)bits file.
But we cannot restore two different nbits files from one single (n  1)bits files, so the compression scheme does not exist.
16.4 Matroids and greedy methods ★
16.41
Show that (S, \(\mathcal{I}_k\)) is a matroid, where S is any finite set and \(\mathcal{I}_k\) is the set of all subsets of S of size at most k, where k ≤ S.
 S is a finite set.
 For any B ∈ \(\mathcal{I}_k\) and A ⊆ B, A ≤ B ≤ k, so A ∈ \(\mathcal{I}_k\).
 If A ∈ \(\mathcal{I}_k\), B ∈ \(\mathcal{I}_k\), and A < B, we know A < B ≤ k. For some x ∈ B  A, A ∪ {x} ≤ k, so A ∪ {x} ∈ \(\mathcal{I}_k\).
16.42 ★
Given an m × n matrix T over some field (such as the reals), show that (S, \(\mathcal{I}\)) is a matroid, where S is the set of columns of T and A ∈ \(\mathcal{I}\) if and only if the columns in A are linearly independent.

S is a finite set.

For any B ∈ \(\mathcal{I}\) and A ⊆ B, we know columns in B are linearly independent, since columns of any subsets of B are also linearly independent, so A ∈ \(\mathcal{I}\).

If A ∈ \(\mathcal{I}\), B ∈ \(\mathcal{I}\), and A < B, there are three cases:

If A = ∅, then for any x ∈ B  A, A ∪ {x} = {x}, which is linearly independent, that is: A ∪ {x} ∈ \(\mathcal{I}\).

If A ≠ ∅ and A ∩ B = ∅, we have B  A = B and B ≥ 2. Since columns in B are linearly independent, there must be at least one column in B that is linearly independent with columns in A.
This can be proved by contradiction: If all columns in B are linearly dependent with columns in A, then columns in B will be linearly dependent, which is impossible, so there must be at least one column in B that is linearly independent with columns in A.
Choose an column x in B that is linearly independent with columns in A, we have A ∪ {x} ∈ \(\mathcal{I}\).

If A ≠ ∅ and A ∩ B ≠ ∅, we know columns in B  A are linearly independent with columns in A ∩ B, which means columns in B  A are linearly independent with columns in A. For any x in B  A, we have A ∪ {x} ∈ \(\mathcal{I}\).

16.43 ★
Show that if (S, \(\mathcal{I}\)) is a matroid, then (S, \(\mathcal{I}'\)) is a matroid, where
\(\mathcal{I}'\) = {A′ : S  A′ contains some maximal A ∈ \(\mathcal{I}\)}.
That is, the maximal independent sets of (S, \(\mathcal{I}'\)) are just the complements of the maximal independent sets of (S, \(\mathcal{I}\)).
 S is a finite set.
 For any B ∈ \(\mathcal{I}'\) and A ⊆ B, we know that S  B contains some maximal C ∈ \(\mathcal{I}\), and S  B ⊆ S  A, so S  A also contains C, which means A ∈ \(\mathcal{I}'\).
 Skipped.
16.44 ★
Let S be a finite set and let \(S_1\), \(S_2\), …, \(S_k\) be a partition of S into nonempty disjoint subsets. Define the structure (S, \(\mathcal{I}\)) by the condition that \(\mathcal{I}\) = {A : A ∩ \(S_i\) ≤ 1 for i = 1, 2, …, k}. Show that (S, \(\mathcal{I}\)) is a matroid. That is, the set of all sets A that contain at most one member of each subset in the partition determines the independent sets of a matroid.
 S is a finite set.
 For any B ∈ \(\mathcal{I}\) and A ⊆ B, for any 1 ≤ i ≤ k, A ∩ \(S_i\) = (A ∩ B) ∩ \(S_i\) = A ∩ (B ∩ \(S_i\)). Since B ∩ \(S_i\) ≤ 1, A ∩ \(S_i\) ≤ 1. So A ∈ \(\mathcal{I}\).
 If A ∈ \(\mathcal{I}\), B ∈ \(\mathcal{I}\), and A < B, then B must intersect with more partitions than A. Let x be an element in a partition that B intersects with, but A does not, we have A ∪ {x} ∈ \(\mathcal{I}_k\).
16.45
Show how to transform the weight function of a weighted matroid problem, where the desired optimal solution is a minimumweight maximal independent subset, to make it a standard weightedmatroid problem. Argue carefully that your transformation is correct.
Let the weights in the minimumweight problem be \(w_1\), \(w_2\), … \(w_n\), and the maximum weight be \(w_k\), we can solve transform this problem into a maximumweight problem with weights \(w_k\) + 1  \(w_1\), \(w_k\) + 1  \(w_2\), …, \(w_k\) + 1  \(w_n\).
Proof:
Skipped.
16.5 A taskscheduling problem as a matroid
16.51
Solve the instance of the scheduling problem given in Figure 16.7, but with each penalty \(w_i\) replaced by 80  \(w_i\).
The schedule is:
\(a_5\), \(a_4\), \(a_6\), \(a_3\), \(a_7\), \(a_2\), \(a_1\).
Solution is implemented here.
16.52
Show how to use property 2 of Lemma 16.12 to determine in time O(A) whether or not a given set A of tasks is independent.
Solution is implemented here.
16.X Problems
161 Coin changing
Consider the problem of making change for n cents using the fewest number of coins. Assume that each coin’s value is an integer.
 a. Describe a greedy algorithm to make change consisting of quarters, dimes, nickels, and pennies. Prove that your algorithm yields an optimal solution.
 b. Suppose that the available coins are in the denominations that are powers of c, i.e., the denominations are \(c^0\), \(c_1\), …, \(c^k\) for some integers c > 1 and k ≥ 1. Show that the greedy algorithm always yields an optimal solution.
 c. Give a set of coin denominations for which the greedy algorithm does not yield an optimal solution. Your set should include a penny so that there is a solution for every value of n.
 d. Give an O(n k)time algorithm that makes change for any set of k different coin denominations, assuming that one of the coins is a penny.

a. Solution is implemented here.
Skipped.

b. Skipped.

c. Let the set of coin denominations be {1, 3, 4}. Let the amount of money be 6, greedy algorithm yields a solution of 4 + 1 + 1, while the optimal solution is 3 + 3.

d. Solution is implemented here.
162 Scheduling to minimize average completion time
Suppose you are given a set S = {\(a_1\), \(a_2\), …, \(a_n\)} of tasks, where task \(a_i\) requires \(p_i\) units of processing time to complete, once it has started. You have one computer on which to run these tasks, and the computer can run only one task at a time. Let \(c_i\) be the completion time of task \(a_i\), that is, the time at which task \(a_i\) completes processing. Your goal is to minimize the average completion time, that is, to minimize (1 / n) \(∑_{i = 1}^n c_i\). For example, suppose there are two tasks, \(a_1\) and \(a_2\), with \(p_1\) = 3 and \(p_2\) = 5, and consider the schedule in which \(a_2\) runs first, followed by \(a_1\). Then \(c_2\) = 5, \(c_1\) = 8, and the average completion time is (5 + 8) / 2 = 6.5. If task \(a_1\) runs first, however, then \(c_1\) = 3, \(c_2\) = 8, and the average completion time is (3 + 8) / 2 = 5.5.
 a. Give an algorithm that schedules the tasks so as to minimize the average completion time. Each task must run nonpreemptively, that is, once task \(a_i\) starts, it must run continuously for \(p_i\) units of time. Prove that your algorithm minimizes the average completion time, and state the running time of your algorithm.
 b. Suppose now that the tasks are not all available at once. That is, each task cannot start until its release time \(r_i\). Suppose also that we allow preemption, so that a task can be suspended and restarted at a later time. For example, a task \(a_i\) with processing time \(p_i\) = 6 and release time \(r_i\) = 1 might start running at time 1 and be preempted at time 4. It might then resume at time 10 but be preempted at time 11, and it might finally resume at time 13 and complete at time 15. Task \(a_i\) has run for a total of 6 time units, but its running time has been divided into three pieces. In this scenario, \(a_i\)’s completion time is 15. Give an algorithm that schedules the tasks so as to minimize the average completion time in this new scenario. Prove that your algorithm minimizes the average completion time, and state the running time of your algorithm.

a. Solution is implemented here.
The idea is process the task with least processing time first. We can prove this strategy minimizes the average completion time, we can show that if there is a pair of tasks in which the one with longer processing time gets scheduled first, then we can get a better average completion time by swap the schedule order of these two tasks.
Assume we have a schedule \(a_{s_1}\), \(a_{s_2}\), …, \(a_{s_n}\), in which there is a pair of tasks \(a_{s_i}\) and \(a_{s_j}\), we have i < j but \(p_{s_i}\) > \(p_{s_j}\).
We split the schedule into five parts:
 Tasks \(a_{s_1}\), \(a_{s_2}\), …, \(a_{s_{i  1}}\),
 Task \(a_{s_i}\),
 Tasks \(a_{s_{i + 1}}\), \(a_{s_{i + 2}}\), …, \(a_{s_{j  1}}\),
 Task \(a_{s_j}\),
 Tasks \(a_{s_{j + 1}}\), \(a_{s_{j + 2}}\), …, \(a_n\).
If we swap task \(a_{s_i}\) with task \(a_{s_j}\) in the schedule, then the completion time in part 1, part 4 and part 5 will stay the same, but the completion time in part 2 and part 3 will be less. So the total completion will be the less, which means the average completion time will be less.
So we can only get the best strategy if all tasks are sorted in increasing processing time order.

b. Solution is implemented here.
I am not sure whether my algorithm is correct, proof is skipped.
The running time is O(n lg n).
163 Acyclic subgraphs
 a. The incidence matrix for an undirected graph G = (V, E) is a V × E matrix M such that \(M_{v e}\) = 1 if edge e is incident on vertex v, and \(M_{v e}\) = 0 otherwise. Argue that a set of columns of M is linearly independent over the field of integers modulo 2 if and only if the corresponding set of edges is acyclic.
 b. Suppose that we associate a nonnegative weight w(e) with each edge in an undirected graph G = (V, E). Give an efficient algorithm to find an acyclic subset of E of maximum total weight.
 c. Let G(V, E) be an arbitrary directed graph, and let (E, \(\mathcal{I}\)) be defined so that A ∈ \(\mathcal{I}\) if and only if A does not contain any directed cycles. Give an example of a directed graph G such that the associated system (E, \(\mathcal{I}\)) is not a matroid. Specify which defining condition for a matroid fails to hold.
 d. The incidence matrix for a directed graph G = (V, E) with no selfloops is a V × E matrix M such that \(M_{v e}\) = 1 if edge e leaves vertex v, \(M_{v e}\) = 1 if edge e enters vertex v, and \(M_{v e}\) = 0 otherwise. Argue that if a set of columns of M is linearly independent, then the corresponding set of edges does not contain a directed cycle.
 e. Exercise 16.42 tells us that the set of linearly independent sets of columns of any matrix M forms a matroid. Explain carefully why the results of parts (c) and (d) are not contradictory. How can there fail to be a perfect correspondence between the notion of a set of edges being acyclic and the notion of the associated set of columns of the incidence matrix being linearly independent?
Skipped.
164 Scheduling variations
Consider the following algorithm for the problem from Section 16.5 of scheduling unittime tasks with deadlines and penalties. Let all n time slots be initially empty, where time slot i is the unitlength slot of time that finishes at time i. We consider the tasks in order of monotonically decreasing penalty. When considering task \(a_j\), if there exists a time slot at or before \(a_j\)’s deadline \(d_j\) that is still empty, assign \(a_j\) to the latest such slot, filling it. If there is no such slot, assign task \(a_j\) to the latest of the as yet unfilled slots.
 a. Argue that this algorithm always gives an optimal answer.
 b. Use the fast disjointset forest presented in Section 21.3 to implement the algorithm efficiently. Assume that the set of input tasks has already been sorted into monotonically decreasing order by penalty. Analyze the running time of your implementation.
Skipped.
165 Offline caching
Modern computers use a cache to store a small amount of data in a fast memory. Even though a program may access large amounts of data, by storing a small subset of the main memory in the cache—a small but faster memory—overall access time can greatly decrease. When a computer program executes, it makes a sequence ⟨\(r_1\), \(r_2\), …, \(r_n\)⟩ of n memory requests, where each request is for a particular data element. For example, a program that accesses 4 distinct elements {a, b, c, d} might make the sequence of requests ⟨d, b, d, b, d, a, c, d, b, a, c, b⟩. Let k be the size of the cache. When the cache contains k elements and the program requests the (k + 1)st element, the system must decide, for this and each subsequent request, which k elements to keep in the cache. More precisely, for each request \(r_i\), the cachemanagement algorithm checks whether element \(r_i\) is already in the cache. If it is, then we have a cache hit; otherwise, we have a cache miss. Upon a cache miss, the system retrieves \(r_i\) from the main memory, and the cachemanagement algorithm must decide whether to keep \(r_i\) in the cache. If it decides to keep \(r_i\) and the cache already holds k elements, then it must evict one element to make room for \(r_i\). The cachemanagement algorithm evicts data with the goal of minimizing the number of cache misses over the entire sequence of requests.
Typically, caching is an online problem. That is, we have to make decisions about which data to keep in the cache without knowing the future requests. Here, however, we consider the offline version of this problem, in which we are given in advance the entire sequence of n requests and the cache size k, and we wish to minimize the total number of cache misses.
We can solve this offline problem by a greedy strategy called furthestinfuture, which chooses to evict the item in the cache whose next access in the request sequence comes furthest in the future.
 a. Write pseudocode for a cache manager that uses the furthestinfuture strategy. The input should be a sequence ⟨\(r_1\), \(r_2\), …, \(r_n\)⟩ of requests and a cache size k, and the output should be a sequence of decisions about which data element (if any) to evict upon each request. What is the running time of your algorithm?
 b. Show that the offline caching problem exhibits optimal substructure.
 c. Prove that furthestinfuture produces the minimum possible number of cache misses.

a. Solution is implemented here.
For test purpose, I used tree based data structure, and the running time is O(n k lg m), where m is the number of memory locations to access.
The running time could be optimized to O(n k) if I use hash table based data structure.

b. Let P(C, i) be the number of minimal cache misses, where C is the set of elements the cache contains, k is the cache size, and i is the number of requests already been processed, we have:
P(C, i) =
 P(C, i + 1), if \(r_{i + 1}\) ∈ C,
 1 + \(\min_{r ∈ C} P(C ∪ \left\lbrace r_{i + 1}\right\rbrace ∖ \left\lbrace r \right\rbrace, i + 1)\), if \(r_{i + 1}\) ∈ C,

c. Skipped.
17 Amortized Analysis
17.1 Aggregate analysis
17.11
If the set of stack operations included a Multipush operation, which pushes k items onto the stack, would the O(1) bound on the amortized cost of stack operations continue to hold?
No. If we do Multipush every time, the amortized cost of each operation will be Θ(k).
17.12
Show that if a Decrement operation were included in the kbit counter example, n operations could cost as much as Θ(n k) time.
If all bits except for the last bit of the counter is 0, while the last bit is 1, then interleave Decrement and Increment operations, then each operation will flip k bits, so n operations will cost Θ(n k) time.
17.13
Suppose we perform a sequence of n operations on a data structure in which the ith operation costs i if i is an exact power of 2, and 1 otherwise. Use aggregate analysis to determine the amortized cost per operation.
The total cost is:
\(∑_{i = 0}^{⌊\lg n⌋} 2^i\) + (n  ⌊lg n⌋  1)
= \(2^{⌊\lg n⌋ + 1}\)  1 + (n  ⌊lg n⌋  1)
≤ \(2^{\lg n + 1}\)  1 + (n  ⌊lg n⌋  1)
= 2 n  1 + (n  ⌊lg n⌋  1)
= 3 n  ⌊lg n⌋  2
So the cost per operation is: O((3 n  ⌊lg n⌋  2) / n) = O(1).
17.2 The accounting method
17.21
Suppose we perform a sequence of stack operations on a stack whose size never exceeds k. After every k operations, we make a copy of the entire stack for backup purposes. Show that the cost of n stack operations, including copying the stack, is O(n) by assigning suitable amortized costs to the various stack operations.
Actual costs:
 Push: 1
 Pop: 1
 Copy: s
Amortized costs:
 Push: 2
 Pop: 2
 Copy: 0
In this way, every Push and Pop operations produces 1 credit, so after every k operations, we have k extra credits, which can be pay for the following Copy operation.
17.22
Redo Exercise 17.13 using an accounting method of analysis.
Actual costs:
 ith operation if i is an exact power of 2: i
 ith operation if i is not an exact power of 2: 1
Amortized costs:
 ith operation if i is an exact power of 2: 2
 ith operation if i is not an exact power of 2: 3
We can save 2 credits for each nonexactpowerof2 operation, and for each exactpowerof2 operation k, there are k / 2  1 consecutive nonexactpowerof2 operations, which can build up to k  2 credits, adding the extra 2 credits from the kth operation, the total number of credits is k, which is enough to pay for the kth operation.
17.23
Suppose we wish not only to increment a counter but also to reset it to zero (i.e., make all bits in it 0). Counting the time to examine or modify a bit as Θ(1), show how to implement a counter as an array of bits so that any sequence of n Increment and Reset operations takes time O(n) on an initially zero counter.
(Hint: Keep a pointer to the highorder 1.)
Like the hint says, we keep track of the highest 1 bit, so Rest operation only need to check partial of the bits.
Assign the following amortized costs:
 Increment when highest 1 bit is changed: 3
 Increment when highest 1 bit is not changed: 2
 Reset: 0
At each step, the remaining credit is the number of 1 bits in the counter plus the location of the highest 1 bit, so it is enough for either Reset or Increment operation.
17.3 The potential method
17.31
Suppose we have a potential function Φ such that Φ(\(D_i\)) ≥ Φ(\(D_0\)) for all i, but Φ(\(D_0\)) ≠ 0. Show that there exists a potential function Φ′ such that Φ′(\(D_0\)) = 0, Φ′(\(D_i\)) ≥ 0 for all i ≥ 1, and the amortized costs using Φ′ are the same as the amortized costs using Φ.
Let Φ′(\(D_i\)) = Φ(\(D_i\))  Φ(\(D_0\)).
17.32
Redo Exercise 17.13 using a potential method of analysis.
Skipped.
17.33
Consider an ordinary binary minheap data structure with n elements supporting the instructions Insert and ExtractMin in O(lg n) worstcase time. Give a potential function Φ such that the amortized cost of Insert is O(lg n) and the amortized cost of ExtractMin is O(1), and show that it works.
Skipped.
17.34
What is the total cost of executing n of the stack operations Push, Pop, and Multipop, assuming that the stack begins with \(s_0\) objects and finishes with \(s_n\) objects?
Let Φ(\(D_i\)) the the number of elements after i operations, we have:
The amortized cost of different operations are:
 Push: 1 + 1 = 2
 Pop: 1 + (1) = 0
 Push: k + (k) = 0
Let p be the number of Push operations, we have:
\(∑_{i = 1}^n c_i\)
= \(∑_{i = 1}^n \hat{c}_i\)  (Φ(\(D_i\))  Φ(\(D_0\)))
= 2 p  \(s_n\) + \(s_0\).
17.35
Suppose that a counter begins at a number with b 1s in its binary representation, rather than at 0. Show that the cost of performing n Increment operations is O(n) if n = Ω(b). (Do not assume that b is constant.)
Let Φ(\(D_i\)) the the number of 1s after i operations, we have:
The amortized cost of Increment is 2.
\(∑_{i = 1}^n c_i\)
= \(∑_{i = 1}^n \hat{c}_i\)  (Φ(\(D_i\))  Φ(\(D_0\)))
= 2 n  \(s_n\) + b
= O(2 n  \(s_n\)) (Since n = Ω(b))
= O(n).
17.36
Show how to implement a queue with two ordinary stacks (Exercise 10.16) so that the amortized cost of each Enqueue and each Dequeue operation is O(1).
I have already implemented the queue in exercise 10.16, see here.
Now I will analyze the running time.
The queue is implemented with two stacks. One is for the queue front, one is for the queue back.
 Enqueue is done by pushing into the back stack.
 Dequeue is done by first checking whether the front stack is empty, if the front stack is empty, pop all items from the back stack and push them into the front stack. Then, pop from the front stack.
Let Φ(\(D_i\)) be the number of elements of the back stack after i operations, we have:
The amortized cost of each operation is:
 Enqueue: 1 + 1 = 2
 Dequeue when front stack has elements: 1 + 0 = 1
 Dequeue when front stack is empty: (k + 1) + (k) = 1
Assume initially, the queue is empty, the total actual cost after n operations is:
\(∑_{i = 1}^n c_i\)
= \(∑_{i = 1}^n \hat{c}_i\)  (Φ(\(D_i\))  Φ(\(D_0\)))
≤ 2 n  \(s_n\)
= O(n).
17.37
Design a data structure to support the following two operations for a dynamic multiset S of integers, which allows duplicate values:
Insert(S, x) inserts x into S.
DeleteLargerHalf(S) deletes the largest ⌈S / 2⌉ elements from S.
Explain how to implement this data structure so that any sequence of m Insert and DeleteLargerHalf operations runs in O(m) time. Your implementation should also include a way to output the elements of S in O(S) time.
Solution is implemented here.
Let Φ(\(D_i\)) be twice the number of elements in the set after i operations, we have:
The amortized cost of each operation is:
 Insert: 1 + 2 = 3
 DeleteLargerHalf: k + 1  2 (⌈k / 2⌉) <= 1, where k is the number of elements in the set
\(∑_{i = 1}^m c_i\)
= \(∑_{i = 1}^m \hat{c}_i\)  (Φ(\(D_i\))  Φ(\(D_0\)))
≤ 3 m  \(s_m\)
= O(m).
17.4 Dynamic tables
17.41
Suppose that we wish to implement a dynamic, openaddress hash table. Why might we consider the table to be full when its load factor reaches some value α that is strictly less than 1? Describe briefly how to make insertion into a dynamic, openaddress hash table run in such a way that the expected value of the amortized cost per insertion is O(1). Why is the expected value of the actual cost per insertion not necessarily O(1) for all insertions?
Skipped.
17.42
Show that if \(\alpha_{i  1}\) ≥ 1 / 2 and the ith operation on a dynamic table is TableDelete, then the amortized cost of the operation with respect to the potential function (17.6) is bounded above by a constant.
Let \(s_i\) be the size of the table after the ith operation, and \(n_i\) be the number of elements after the ith operation.

If \(\alpha_{i  1}\) = 1 / 2, we have \(s_{i  1}\) = 2 \(n_{i  1}\).

If the ith operation triggers a contraction, we have:
\(n_{i  1}\)  1 < \(s_{i  1}\) / 4
⇔ \(n_{i  1}\)  1 < \(n_{i  1}\) / 2
⇔ \(n_{i  1}\) < 2
⇔ \(n_{i  1}\) = 1. (\(n_{i  1}\) ≠ 0 because we will do a deletion on \(n_{i  1}\) elements)Doing a deletion on a constant size dynamic table costs constant time.

If the ith operation does not trigger a contraction, we have:
\(\hat{c}_i\)
= \(c_i\) + Φ(\(D_{i}\))  Φ(\(D_{i  1}\))
= 1 + 0  (\(s_{i  1}\) / 2  (\(n_{i  1}\)  1))
= 1 + 0  (\(n_{i  1}\)  (\(n_{i  1}\)  1))
= 1 + 0  1
= 0.


If \(\alpha_{i  1}\) > 1 / 2, we know that the ith operation does not trigger a contraction.
\(\hat{c}_i\)
= \(c_i\) + Φ(\(D_{i}\))  Φ(\(D_{i  1}\))
= 1 + (2 (\(n_{i  1}\)  1)  \(s_{i  1}\))  (2 \(n_{i  1}\)  \(s_{i  1}\))
= 1 + 0  (\(n_{i  1}\)  (\(n_{i  1}\)  1))
= 1.
17.43
Suppose that instead of contracting a table by halving its size when its load factor drops below 1 / 4, we contract it by multiplying its size by 2 / 3 when its load factor drops below 1 / 3. Using the potential function
Φ(T) = 2 ⋅ T.num  T.size,
show that the amortized cost of a TableDelete that uses this strategy is bounded above by a constant.
Skipped.
17.X Problems
171 Bitreversed binary counter
Chapter 30 examines an important algorithm called the fast Fourier transform, or FFT. The first step of the FFT algorithm performs a bitreversal permutation on an input array A[0‥n  1] whose length is n = \(2^k\) for some nonnegative integer k. This permutation swaps elements whose indices have binary representations that are the reverse of each other.
We can express each index a as a kbit sequence ⟨\(a_{k  1}\), \(a_{k  2}\), …, \(a_0\)⟩, where a = \(∑_{i = 0}^{k  1} a_i 2^i\). We define
\(\text{rev}_k\)(⟨\(a_{k  1}\), \(a_{k  2}\), …, \(a_0\)⟩) = ⟨\(a_0\), \(a_1\), …, \(a_{k  1}\)⟩;
thus,
\(\text{rev}_k\)(a) = \(\displaystyle ∑_{i = 0}^{k  1} a_{k  i  1} 2^i\).
For example, if n = 16 (or, equivalently, k = 4), then \(\text{rev}_k\)(3) = 12, since the 4bit representation of 3 is 0011, which when reversed gives 1100, the 4bit representation of 12.
 a. Given a function \(\text{rev}_k\) that runs in Θ(k) time, write an algorithm to perform the bitreversal permutation on an array of length n = \(2^k\) in O(n k) time.
We can use an algorithm based on an amortized analysis to improve the running time of the bitreversal permutation. We maintain a “bitreversed counter” and a procedure BitReversedIncrement that, when given a bitreversedcounter value a, produces \(\text{rev}_k\)(\(\text{rev}_k\)(a) + 1). If k = 4, for example, and the bitreversed counter starts at 0, then successive calls to BitReversedIncrement produce the sequence
0000, 1000, 0100, 1100, 0010, 1010, … = 0, 8, 4, 12, 2, 10, ….
 b. Assume that the words in your computer store kbit values and that in unit time, your computer can manipulate the binary values with operations such as shifting left or right by arbitrary amounts, bitwiseAND, bitwiseOR, etc. Describe an implementation of the BitReversedIncrement procedure that allows the bitreversal permutation on an nelement array to be performed in a total of O(n) time.
 c. Suppose that you can shift a word left or right by only one bit in unit time. Is it still possible to implement an O(n)time bitreversal permutation?
Solution is implemented here.
172 Making binary search dynamic
Binary search of a sorted array takes logarithmic search time, but the time to insert a new element is linear in the size of the array. We can improve the time for insertion by keeping several sorted arrays.
Specifically, suppose that we wish to support Search and Insert on a set of n elements. Let k = ⌈lg(n + 1)⌉, and let the binary representation of n be ⟨\(n_{k  1}\), \(n_{k  2}\), …, \(n_0\)⟩. We have k sorted arrays \(A_0\), \(A_1\), …, \(A_{k  1}\), where for i = 0, 1, …, k  1, the length of array \(A_i\) is \(2^i\). Each array is either full or empty, depending on whether \(n_i\) = 1 or \(n_i\) = 0, respectively. The total number of elements held in all k arrays is therefore \(∑_{i = 0}^{k  1} n_i 2^i\) = n. Although each individual array is sorted, elements in different arrays bear no particular relationship to each other.
 a. Describe how to perform the Search operation for this data structure. Analyze its worstcase running time.
 b. Describe how to perform the Insert operation. Analyze its worstcase and amortized running times.
 c. Discuss how to implement Delete.
Solution is implemented here.

a. If the set has n = \(2^k\) elements, and the element to search is in the last slot, then the worst case search time is:
\(∑_{i = 1}^k \lg (2^i)\)
= \(∑_{i = 1}^k i\)
= k (k + 1) / 2. 
b.
Worst case running time occurs when a new element is being inserted into a set with n = \(2^k\)  1 elements. The running time is Θ(n).
Skipped amortized running time analysis.
173 Amortized weightbalanced trees
Consider an ordinary binary search tree augmented by adding to each node x the attribute x.size giving the number of keys stored in the subtree rooted at x. Let α be a constant in the range 1 / 2 ≤ α < 1. We say that a given node x is αbalanced if x.left.size ≤ α ⋅ x.size and x.right.size ≤ α ⋅ x.size. The tree as a whole is αbalanced if every node in the tree is αbalanced. The following amortized approach to maintaining weightbalanced trees was suggested by G. Varghese.
 a. A 1 / 2balanced tree is, in a sense, as balanced as it can be. Given a node x in an arbitrary binary search tree, show how to rebuild the subtree rooted at x so that it becomes 1 / 2balanced. Your algorithm should run in time Θ(x.size), and it can use O(x.size) auxiliary storage.
 b. Show that performing a search in an nnode αbalanced binary search tree takes O(lg n) worstcase time.
For the remainder of this problem, assume that the constant α is strictly greater than 1 / 2. Suppose that we implement Insert and Delete as usual for an nnode binary search tree, except that after every such operation, if any node in the tree is no longer αbalanced, then we “rebuild” the subtree rooted at the highest such node in the tree so that it becomes 1 / 2balanced.
We shall analyze this rebuilding scheme using the potential method. For a node x in a binary search tree T, we define
Δ(x) = x.left.size  x.right.size,
and we define the potential of T as
Φ(T) = c \(\displaystyle ∑_{x ∈ T: Δ(x) ≥ 2} Δ(x)\),
where c is a sufficiently large constant that depends on α.
 c. Argue that any binary search tree has nonnegative potential and that a 1 / 2balanced tree has potential 0.
 d. Suppose that m units of potential can pay for rebuilding an mnode subtree. How large must c be in terms of α in order for it to take O(1) amortized time to rebuild a subtree that is not αbalanced?
 e. Show that inserting a node into or deleting a node from an nnode αbalanced tree costs O(lg n) amortized time.
The tree is implemented here.
Skipped running time analysis.
174 The cost of restructuring redblack trees
There are four basic operations on redblack trees that perform structural modifications: node insertions, node deletions, rotations, and color changes. We have seen that RBInsert and RBDelete use only O(1) rotations, node insertions, and node deletions to maintain the redblack properties, but they may make many more color changes.
 a. Describe a legal redblack tree with n nodes such that calling RBInsert to add the (n + 1)st node causes Ω(lg n) color changes. Then describe a legal redblack tree with n nodes for which calling RBDelete on a particular node causes Ω(lg n) color changes.
Although the worstcase number of color changes per operation can be logarithmic, we shall prove that any sequence of m RBInsert and RBDelete operations on an initially empty redblack tree causes O(m) structural modifications in the worst case. Note that we count each color change as a structural modification.
b. Some of the cases handled by the main loop of the code of both RBInsertFixup and RBDeleteFixup are terminating: once encountered, they cause the loop to terminate after a constant number of additional operations. For each of the cases of RBInsertFixup and RBDeleteFixup, specify which are terminating and which are not. (Hint: Look at Figures 13.5, 13.6, and 13.7.)
We shall first analyze the structural modifications when only insertions are performed. Let T be a redblack tree, and define Φ(T) to be the number of red nodes in T. Assume that 1 unit of potential can pay for the structural modifications performed by any of the three cases of RBInsertFixup.
c. Let T′ be the result of applying Case 1 of RBInsertFixup to T. Argue that Φ(T′) = Φ(T)  1.
d. When we insert a node into a redblack tree using RBInsert, we can break the operation into three parts. List the structural modifications and potential changes resulting from lines 1–16 of RBInsert, from nonterminating cases of RBInsertFixup, and from terminating cases of RBInsertFixup.
e. Using part (d), argue that the amortized number of structural modifications performed by any call of RBInsert is O(1).
We now wish to prove that there are O(m) structural modifications when there are both insertions and deletions. Let us define, for each node x,
w(x) = \(\begin{cases} 0 &\text{if $x$ is red,}\\ 1 &\text{if $x$ is black and has no red children,}\\ 0 &\text{if $x$ is black and has one red child,}\\ 2 &\text{if $x$ is black and has two red children.} \end{cases}\)
Now we redefine the potential of a redblack tree T as
Φ(T) = \(∑_{x ∈ T} w(x)\),
and let T′ be the tree that results from applying any nonterminating case of RBInsertFixup or RBDeleteFixup to T.
 f. Show that Φ(T′) ≤ Φ(T)  1 for all nonterminating cases of RBInsertFixup. Argue that the amortized number of structural modifications performed by any call of RBInsertFixup is O(1).
 g. Show that Φ(T′) ≤ Φ(T)  1 for all nonterminating cases of RBDeleteFixup. Argue that the amortized number of structural modifications performed by any call of RBDeleteFixup is O(1).
 h. Complete the proof that in the worst case, any sequence of m RBInsert and RBDelete operations performs O(m) structural modifications.
Skipped.
175 Competitive analysis of selforganizing lists with movetofront
A selforganizing list is a linked list of n elements, in which each element has a unique key. When we search for an element in the list, we are given a key, and we want to find an element with that key.
A selforganizing list has two important properties:
 To find an element in the list, given its key, we must traverse the list from the beginning until we encounter the element with the given key. If that element is the kth element from the start of the list, then the cost to find the element is k.
 We may reorder the list elements after any operation, according to a given rule with a given cost. We may choose any heuristic we like to decide how to reorder the list.
Assume that we start with a given list of n elements, and we are given an access sequence σ = ⟨\(\sigma_1\), \(\sigma_2\), …, \(\sigma_m\)⟩ of keys to find, in order. The cost of the sequence is the sum of the costs of the individual accesses in the sequence.
Out of the various possible ways to reorder the list after an operation, this problem focuses on transposing adjacent list elements—switching their positions in the list—with a unit cost for each transpose operation. You will show, by means of a potential function, that a particular heuristic for reordering the list, movetofront, entails a total cost no worse than 4 times that of any other heuristic for maintaining the list order—even if the other heuristic knows the access sequence in advance! We call this type of analysis a competitive analysis.
For a heuristic H and a given initial ordering of the list, denote the access cost of sequence σ by \(C_H\)(σ). Let m be the number of accesses in σ.
 a. Argue that if heuristic H does not know the access sequence in advance, then the worstcase cost for H on an access sequence σ is \(C_H\)(σ) = Ω(m n).
With the movetofront heuristic, immediately after searching for an element x, we move x to the first position on the list (i.e., the front of the list).
Let \(\text{rank}_L\)(x) denote the rank of element x in list L, that is, the position of x in list L. For example, if x is the fourth element in L, then \(\text{rank}_L\)(x) = 4. Let \(c_i\) denote the cost of access \(\sigma_i\) using the movetofront heuristic, which includes the cost of finding the element in the list and the cost of moving it to the front of the list by a series of transpositions of adjacent list elements.
 b. Show that if \(\sigma_i\) accesses element x in list L using the movetofront heuristic, then \(c_i\) = 2 ⋅ \(\text{rank}_L\)(x)  1.
Now we compare movetofront with any other heuristic H that processes an access sequence according to the two properties above. Heuristic H may transpose elements in the list in any way it wants, and it might even know the entire access sequence in advance.
Let \(L_i\) be the list after access \(\sigma_i\) using movetofront, and let \(L_i^*\) be the list after access \(\sigma_i\) using heuristic H. We denote the cost of access \(\sigma_i\) by \(c_i\) for movetofront and by \(c_i^*\) for heuristic H. Suppose that heuristic H performs \(t_i^*\) transpositions during access \(\sigma_i\).
 c. In part (b), you showed that \(c_i\) = 2 ⋅ \(rank_{L_{i  1}}\)(x)  1. Now show that \(c_i^*\) = \(\text{rank}_{L_{i  1}^*}(x) + t_i^*\).
We define an inversion in list \(L_i\) as a pair of elements y and z such that y precedes z in \(L_i\) and z precedes y in list \(L_i^*\). Suppose that list \(L_i\) has \(q_i\) inversions after processing the access sequence ⟨\(\sigma_1\), \(\sigma_2\), …, \(\sigma_i\)⟩. Then, we define a potential function Φ that maps \(L_i\) to a real number by Φ(\(L_i\)) = 2 \(q_i\). For example, if \(L_i\) has the elements ⟨e, c, a, d, b⟩ and \(L_i^*\) has the elements ⟨c, a, b, d, e⟩, then \(L_i\) has 5 inversions ((e, c), (e, a), (e, d), (e, b), (d, b)), and so Φ(\(L_i\)) = 10. Observe that Φ(\(L_i\)) ≥ 0 for all i and that, if movetofront and heuristic H start with the same list \(L_0\), then Φ(\(L_0\)) = 0.
 d. Argue that a transposition either increases the potential by 2 or decreases the potential by 2.
Suppose that access \(\sigma_i\) finds the element x. To understand how the potential changes due to \(\sigma_i\), let us partition the elements other than x into four sets, depending on where they are in the lists just before the ith access:
Set A consists of elements that precede x in both \(L_{i  1}\) and \(L_{i  1}^*\).
Set B consists of elements that precede x in \(L_{i  1}\) and follow x in \(L_{i  1}^*\).
Set C consists of elements that follow x in \(L_{i  1}\) and precede x in \(L_{i  1}^*\).
Set D consists of elements that follow x in both \(L_{i  1}\) and \(L_{i  1}^*\).
e. Argue that \(\text{rank}_{L_{i  1}}\)(x) = A + B + 1 and \(\text{rank}_{L_{i  1}^*}\)(x) = A + C + 1.
f. Show that access \(\sigma_i\) causes a change in potential of
Φ(\(L_i\))  Φ(\(L_{i  1}\)) ≤ 2 (A  B + \(t_i^*\)),
where, as before, heuristic H performs \(t_i^*\) transpositions during access \(\sigma_i\).
Define the amortized cost \(\hat{c}_i\) of access \(\sigma_i\) by \(\hat{c}_i\) = \(c_i\) + Φ(\(L_i\))  Φ(\(L_{i  1}\)).
g. Show that the amortized cost \(\hat{c}_i\) of access \(\sigma_i\) is bounded from above by 4 \(c_i^*\).
h. Conclude that the cost \(C_\text{MTF}\)(σ) of access sequence σ with movetofront is at most 4 times the cost \(C_H\)(σ) of σ with any other heuristic H, assuming that both heuristics start with the same list.
Skipped.
V Advanced Data Structures
18 BTrees
18.1 Definition of Btrees
18.11
Why don’t we allow a minimum degree of t = 1?
If t = 1, every node could have either zero or one key, so the tree is essentially a linked list, which does not perform well.
18.12
For what values of t is the tree of Figure 18.1 a legal Btree?
Solving
\(\begin{cases} t  1 ≤ 2 ≤ 2 t  1\\ t  1 ≤ 3 ≤ 2 t  1 \end{cases}\),
We have:
2 ≤ t ≤ 3. So t could be 2 or 3.
18.13
Show all legal Btrees of minimum degree 2 that represent {1, 2, 3, 4, 5}.
Every nodes can have 1, 2 or 3 keys.
When root have 1 keys:
┌───┐
│ 2 │
└─┬─┘
┌───┴────┐
┌─┴─┐ ┌───┴───┐
│ 1 │ │ 3 4 5 │
└───┘ └───────┘
┌───┐
│ 3 │
└─┬─┘
┌───┴────┐
┌──┴──┐ ┌──┴──┐
│ 1 2 │ │ 4 5 │
└─────┘ └─────┘
┌───┐
│ 4 │
└─┬─┘
┌───┴────┐
┌───┴───┐ ┌─┴─┐
│ 1 2 3 │ │ 5 │
└───────┘ └───┘
When root have 2 keys:
┌─────┐
│ 2 4 │
└──┬──┘
┌──────┼──────┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐
│ 1 │ │ 3 │ │ 5 │
└───┘ └───┘ └───┘
When root have 3 keys, it has 4 children, which requires at least 7 keys, so it is not possible.
18.14
As a function of the minimum degree t, what is the maximum number of keys that can be stored in a Btree of height h?
(2 t  1) \(∑_{i = 0}^h (2 t)^i\)
= \((2 t)^{h + 1}\)  1.
18.15
Describe the data structure that would result if each black node in a redblack tree were to absorb its red children, incorporating their children with its own.
A Btree with minimum degree 2.
18.2 Basic operations on Btrees
18.21
Show the results of inserting the keys
F, S, Q, K, C, L, H, T, V, W, M, R, N, P, A, B, X, Y, D, Z, E
in order into an empty Btree with minimum degree 2. Draw only the configurations of the tree just before some node must split, and also draw the final configuration.
Solution is proved here.
Inserting K, C, L, H, T, V, W, M, R, N, P, A, B, X, Y, D, Z, E:
┌───────┐
│ F Q S │
└───────┘
Inserting L, H, T, V, W, M, R, N, P, A, B, X, Y, D, Z, E:
┌───┐
│ Q │
└─┬─┘
┌───┴────┐
┌───┴───┐ ┌─┴─┐
│ C F K │ │ S │
└───────┘ └───┘
Inserting W, M, R, N, P, A, B, X, Y, D, Z, E:
┌─────┐
│ F Q │
└──┬──┘
┌────────┬┴─────────┐
┌─┴─┐ ┌───┴───┐ ┌───┴───┐
│ C │ │ H K L │ │ S T V │
└───┘ └───────┘ └───────┘
Inserting M, R, N, P, A, B, X, Y, D, Z, E:
┌───────┐
│ F Q T │
└───┬───┘
┌────────┬───┴────┬───────┐
┌─┴─┐ ┌───┴───┐ ┌─┴─┐ ┌──┴──┐
│ C │ │ H K L │ │ S │ │ V W │
└───┘ └───────┘ └───┘ └─────┘
Inserting M, R, N, P, A, B, X, Y, D, Z, E:
┌───┐
│ Q │
└─┬─┘
┌────────┴────────┐
┌─┴─┐ ┌─┴─┐
│ F │ │ T │
└─┬─┘ └─┬─┘
┌───┴────┐ ┌───┴───┐
┌─┴─┐ ┌───┴───┐ ┌─┴─┐ ┌──┴──┐
│ C │ │ H K L │ │ S │ │ V W │
└───┘ └───────┘ └───┘ └─────┘
Inserting P, A, B, X, Y, D, Z, E:
┌───┐
│ Q │
└─┬─┘
┌──────────┴──────────┐
┌──┴──┐ ┌─┴─┐
│ F K │ │ T │
└──┬──┘ └─┬─┘
┌──────┬┴───────┐ ┌───┴────┐
┌─┴─┐ ┌─┴─┐ ┌───┴───┐ ┌──┴──┐ ┌──┴──┐
│ C │ │ H │ │ L M N │ │ R S │ │ V W │
└───┘ └───┘ └───────┘ └─────┘ └─────┘
Inserting A, B, X, Y, D, Z, E:
┌───┐
│ Q │
└─┬─┘
┌───────────┴───────────┐
┌───┴───┐ ┌─┴─┐
│ F K M │ │ T │
└───┬───┘ └─┬─┘
┌──────┬───┴──┬───────┐ ┌───┴────┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐
│ C │ │ H │ │ L │ │ N P │ │ R S │ │ V W │
└───┘ └───┘ └───┘ └─────┘ └─────┘ └─────┘
Inserting Y, D, Z, E:
┌─────┐
│ K Q │
└──┬──┘
┌───────────────┬┴────────────────┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐
│ F │ │ M │ │ T │
└─┬─┘ └─┬─┘ └─┬─┘
┌───┴────┐ ┌───┴───┐ ┌────┴────┐
┌───┴───┐ ┌─┴─┐ ┌─┴─┐ ┌──┴──┐ ┌──┴──┐ ┌───┴───┐
│ A B C │ │ H │ │ L │ │ N P │ │ R S │ │ V W X │
└───────┘ └───┘ └───┘ └─────┘ └─────┘ └───────┘
Inserting D, Z, E:
┌─────┐
│ K Q │
└──┬──┘
┌───────────────┬┴────────────────┐
┌─┴─┐ ┌─┴─┐ ┌─┴─┐
│ F │ │ M │ │ T │
└─┬─┘ └─┬─┘ └─┬─┘
┌───┴────┐ ┌───┴───┐ ┌────┴────┐
┌───┴───┐ ┌─┴─┐ ┌─┴─┐ ┌──┴──┐ ┌──┴──┐ ┌───┴───┐
│ A B C │ │ H │ │ L │ │ N P │ │ R S │ │ V W X │
└───────┘ └───┘ └───┘ └─────┘ └─────┘ └───────┘
Final:
┌─────┐
│ K Q │
└──┬──┘
┌───────────────────┼────────────────────┐
┌──┴──┐ ┌─┴─┐ ┌──┴──┐
│ B F │ │ M │ │ T W │
└──┬──┘ └─┬─┘ └──┬──┘
┌────────┼────────┐ ┌───┴───┐ ┌───────┼────────┐
┌─┴─┐ ┌───┴───┐ ┌─┴─┐ ┌─┴─┐ ┌──┴──┐ ┌──┴──┐ ┌─┴─┐ ┌───┴───┐
│ A │ │ C D E │ │ H │ │ L │ │ N P │ │ R S │ │ V │ │ X Y Z │
└───┘ └───────┘ └───┘ └───┘ └─────┘ └─────┘ └───┘ └───────┘
18.22
Explain under what circumstances, if any, redundant DiskRead or DiskWrite operations occur during the course of executing a call to BTreeInsert. (A redundant DiskRead is a DiskRead for a page that is already in memory. A redundant DiskWrite writes to disk a page of information that is identical to what is already stored there.)
Skipped.
18.23
Explain how to find the minimum key stored in a Btree and how to find the predecessor of a given key stored in a Btree.
Minimum key is the first key of the the leftmost leaf node.
For the predecessor:
 If the key is in a leaf node:
 If the key is the first key in the leaf node, find its closest ancestor whose ith child contains the key where i > 1, then the predecessor is the maximum key of the (i  1)th child of that ancestor.
 Otherwise the predecessor of the key is the preceding key of the key in the same leaf node.
 Otherwise the key is in an internal node. let i be the index of the key, then the predecessor is the maximum key of the ith child.
18.24 ★
Suppose that we insert the keys {1, 2, …, n} into an empty Btree with minimum degree 2. How many nodes does the final Btree have?
Skipped.
18.25
Since leaf nodes require no pointers to children, they could conceivably use a different (larger) t value than internal nodes for the same disk page size. Show how to modify the procedures for creating and inserting into a Btree to handle this variation.
Skipped.
18.26
Suppose that we were to implement BTreeSearch to use binary search rather than linear search within each node. Show that this change makes the CPU time required O(lg n), independently of how t might be chosen as a function of n.
After using binary search, every node requires O(lg t) time, so the total running time is:
O(lg t \(\log_t n\))
= O(\(\log_t n^{\log_2 t}\))
= O(\(\log_t t^{\log_2 n}\))
= O(\(\log_2 n\))
= O(lg n)
18.27
Suppose that disk hardware allows us to choose the size of a disk page arbitrarily, but that the time it takes to read the disk page is a + b t, where a and b are specified constants and t is the minimum degree for a Btree using pages of the selected size. Describe how to choose t so as to minimize (approximately) the Btree search time. Suggest an optimal value of t for the case in which a = 5 milliseconds and b = 10 microseconds.
Skipped.
18.3 Deleting a key from a Btree
18.31
Show the results of deleting C, P, and V, in order, from the tree of Figure 18.8(f).
After deleting C:
┌─────────┐
│ L P T X │
└────┬────┘
┌──────────┬────────┴┬─────────┬────────┐
┌────┴────┐ ┌──┴──┐ ┌───┴───┐ ┌──┴──┐ ┌──┴──┐
│ A E J K │ │ N O │ │ Q R S │ │ U V │ │ Y Z │
└─────────┘ └─────┘ └───────┘ └─────┘ └─────┘
After deleting P:
┌─────────┐
│ L Q T X │
└────┬────┘
┌──────────┬───────┴┬────────┬────────┐
┌────┴────┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐ ┌──┴──┐
│ A E J K │ │ N O │ │ R S │ │ U V │ │ Y Z │
└─────────┘ └─────┘ └─────┘ └─────┘ └─────┘
After deleting V:
┌───────┐
│ L Q X │
└───┬───┘
┌──────────┬────┴─────┬──────────┐
┌────┴────┐ ┌──┴──┐ ┌────┴────┐ ┌──┴──┐
│ A E J K │ │ N O │ │ R S T U │ │ Y Z │
└─────────┘ └─────┘ └─────────┘ └─────┘
See here for verification.
18.32
Write pseudocode for BTreeDelete.
Real code is implemented here.
18.X Problems
181 Stacks on secondary storage
Consider implementing a stack in a computer that has a relatively small amount of fast primary memory and a relatively large amount of slower disk storage. The operations Push and Pop work on singleword values. The stack we wish to support can grow to be much larger than can fit in memory, and thus most of it must be stored on disk.
A simple, but inefficient, stack implementation keeps the entire stack on disk. We maintain in memory a stack pointer, which is the disk address of the top element on the stack. If the pointer has value p, the top element is the (p mod m)th word on page ⌊p / m⌋ of the disk, where m is the number of words per page. To implement the Push operation, we increment the stack pointer, read the appropriate page into memory from disk, copy the element to be pushed to the appropriate word on the page, and write the page back to disk. A Pop operation is similar. We save the top of the stack, decrement the stack pointer, read in the appropriate page from disk, and return the saved value. We need not write back the page, since it was not modified.
Because disk operations are relatively expensive, we count two costs for any implementation: the total number of disk accesses and the total CPU time. Any disk access to a page of m words incurs charges of one disk access and Θ(m) CPU time.
 a. Asymptotically, what is the worstcase number of disk accesses for n stack operations using this simple implementation? What is the CPU time for n stack operations? (Express your answer in terms of m and n for this and subsequent parts.)
Now consider a stack implementation in which we keep one page of the stack in memory. (We also maintain a small amount of memory to keep track of which page is currently in memory.) We can perform a stack operation only if the relevant disk page resides in memory. If necessary, we can write the page currently in memory to the disk and read in the new page from the disk to memory. If the relevant disk page is already in memory, then no disk accesses are required.
 b. What is the worstcase number of disk accesses required for n Push operations? What is the CPU time?
 c. What is the worstcase number of disk accesses required for n stack operations? What is the CPU time?
Suppose that we now implement the stack by keeping two pages in memory (in addition to a small number of words for bookkeeping).
 d. Describe how to manage the stack pages so that the amortized number of disk accesses for any stack operation is O(1 / m) and the amortized CPU time for any stack operation is O(1).

a. Number of disk accesses is Θ(n), since every operation needs one (Pop) or two (Push) disk accesses. CPU time is Θ(m n).

b. Number of disk accesses is Θ(n / m). CPU time is Θ(n).

c. Number of disk accesses is Θ(n). CPU time is Θ(m n). Worst case happens when Push and Pop on page boundaries. Assume the in memory page is full, the following operation sequences will trigger a worst case:
⟨Push, Pop, Pop, Push, Push, Pop, Pop, Push, Push …⟩

d. Let the two pages in memory pages be bottom page and top page.
 Push:
 If the in memory pages are not full, the operation is purely in memory. No disk access is required.
 If the in memory pages are full, flush the bottom in memory page to disk, swap top page and bottom page. (That is, making top page bottom page, and top page empty.)
 Pop:
 If the in memory pages are not empty, the operation is purely in memory. No disk access is required.
 If the in memory pages are empty, read one page from the disk as bottom page.
 Push:
182 Joining and splitting 234 trees
The join operation takes two dynamic sets S′ and S′′ and an element x such that for any x′ ∈ S′ and x′′ ∈ S′′, we have x′.key < x.key < x′′.key. It returns a set S = S′ ∪ {x} ∪ S′′. The split operation is like an “inverse” join: given a dynamic set S and an element x ∈ S, it creates a set S′ that consists of all elements in S  {x} whose keys are less than x.key and a set S′′ that consists of all elements in S  {x} whose keys are greater than x.key. In this problem, we investigate how to implement these operations on 234 trees. We assume for convenience that elements consist only of keys and that all key values are distinct.
 a. Show how to maintain, for every node x of a 234 tree, the height of the subtree rooted at x as an attribute x.height. Make sure that your implementation does not affect the asymptotic running times of searching, insertion, and deletion.
 b. Show how to implement the join operation. Given two 234 trees T′ and T′′ and a key k, the join operation should run in O(1 + h′  h′′) time, where h′ and h′′ are the heights of T′ and T′′, respectively.
 c. Consider the simple path p from the root of a 234 tree T to a given key k, the set S′ of keys in T that are less than k, and the set S′′ of keys in T that are greater than k. Show that p breaks S′ into a set of trees {\(T_0'\), \(T_1'\), …, \(T_m'\)} and a set of keys {\(k_0'\), \(k_1'\), …, \(k_m'\)}, where, for i = 1, 2, …, m, we have y < \(k_i'\) < z for any keys y ∈ \(T_{i  1}'\) and z ∈ \(T_i'\). What is the relationship between the heights of \(T_{i  1}'\) and \(T_i'\)? Describe how p breaks S′′ into sets of trees and keys.
 d. Show how to implement the split operation on T. Use the join operation to assemble the keys in S′ into a single 234 tree T′ and the keys in S′′ into a single 234 tree T′′. The running time of the split operation should be O(lg n), where n is the number of keys in T. (Hint: The costs for joining should telescope.)
Skipped.
Data Structures for Disjoint Sets
21.1 Disjointset operations
21.11
Suppose that ConnectedComponents is run on the undirected graph G = (V, E), where V = {a, b, c, d, e, f, g, h, i, j, k} and the edges of E are processed in the order (d, i), (f, k), (g, i), (b, g), (a, h), (i, j), (d, k), (b, j), (d, f), (g, j), (a, e). List the vertices in each connected component after each iteration of lines 3–5.
Edge processed  Collection of disjoint sets 

(d, i)  {a} {b} {c} {d, i} {e} {f} {g} {h} {j} {k} 
(f, k)  {a} {b} {c} {d, i} {e} {f, k} {g} {h} {j} 
(g, i)  {a} {b} {c} {d, g, i} {e} {f, k} {h} {j} 
(b, g)  {a} {b, d, g, i} {c} {e} {f, k} {h} {j} 
(a, h)  {a, h} {b, d, g, i} {c} {e} {f, k} {j} 
(i, j)  {a, h} {b, d, g, i, j} {c} {e} {f, k} 
(d, k)  {a, h} {b, d, f, g, i, j, k} {c} {e} 
(b, j)  {a, h} {b, d, f, g, i, j, k} {c} {e} 
(d, f)  {a, h} {b, d, f, g, i, j, k} {c} {e} 
(g, j)  {a, h} {b, d, f, g, i, j, k} {c} {e} 
(a, e)  {a, e, h} {b, d, f, g, i, j, k} {c} 
21.12
Show that after all edges are processed by ConnectedComponents, two vertices are in the same connected component if and only if they are in the same set.
Skipped.
21.13
During the execution of ConnectedComponents on an undirected graph G = (V, E) with k connected components, how many times is FindSet called? How many times is Union called? Express your answers in terms of V, E, and k.
FindSet is called 2 E time, Union is called V  k times, since initially, there are V components, and each Union call reduces one connected component.
Linkedlist representation of disjoint sets
21.21
Write pseudocode for MakeSet, FindSet, and Union using the linkedlist representation and the weightedunion heuristic. Make sure to specify the attributes that you assume for set objects and list objects.
Skipped.
21.22
Show the data structure that results and the answers returned by the FindSet operations in the following program. Use the linkedlist representation with the weightedunion heuristic.
 for i = 1 to 16
 MakeSet(\(x_i\))
 for i = 1 to 15 by 2
 Union(\(x_i\), \(x_{i + 1}\))
 for i = 1 to 13 by 4
 Union(\(x_i\), \(x_{i + 2}\))
 Union(\(x_1\), \(x_5\))
 Union(\(x_{11}\), \(x_{13}\))
 Union(\(x_1\), \(x_{10}\))
 FindSet(\(x_2\))
 FindSet(\(x_9\))
Assume that if the sets containing \(x_i\) and \(x_j\) have the same size, then the operation Union(\(x_i\), \(x_j\)}) appends \(x_j\)’s list onto \(x_i\)’s list.
Skipped.
21.23
Adapt the aggregate proof of Theorem 21.1 to obtain amortized time bounds of O(1) for MakeSet and FindSet and O(lg n) for Union using the linkedlist representation and the weightedunion heuristic.
Skipped.
21.24
Give a tight asymptotic bound on the running time of the sequence of operations in Figure 21.3 assuming the linkedlist representation and the weightedunion heuristic.
Skipped.
21.25
Professor Gompers suspects that it might be possible to keep just one pointer in each set object, rather than two (head and tail), while keeping the number of pointers in each list element at two. Show that the professor’s suspicion is well founded by describing how to represent each set by a linked list such that each operation has the same running time as the operations described in this section. Describe also how the operations work. Your scheme should allow for the weightedunion heuristic, with the same effect as described in this section. (Hint: Use the tail of a linked list as its set’s representative.)
Skipped.
21.26
Suggest a simple change to the Union procedure for the linkedlist representation that removes the need to keep the tail pointer to the last object in each list. Whether or not the weightedunion heuristic is used, your change should not change the asymptotic running time of the Union procedure. (Hint: Rather than appending one list to another, splice them together.)
Skipped.
21.3 Disjointset forests
21.31
Redo Exercise 21.22 using a disjointset forest with union by rank and path compression.
Skipped.
21.32
Write a nonrecursive version of FindSet with path compression.
Solution is implemented here.
21.33
Give a sequence of m MakeSet, Union, and FindSet operations, n of which are MakeSet operations, that takes Ω(m lg n) time when we use union by rank only.
Skipped.
21.34
Suppose that we wish to add the operation PrintSet(x), which is given a node x and prints all the members of x’s set, in any order. Show how we can add just a single attribute to each node in a disjointset forest so that PrintSet(x) takes time linear in the number of members of x’s set and the asymptotic running times of the other operations are unchanged. Assume that we can print each member of the set in O(1) time.
Link the nodes in the same set into a circular linked list.
Solution is implemented here.
21.35 ★
Show that any sequence of m MakeSet, FindSet, and Link operations, where all the Link operations appear before any of the FindSet operations, takes only O(m) time if we use both path compression and union by rank. What happens in the same situation if we use only the pathcompression heuristic?
Skipped.
21.X Problems
211 Offline minimum
The offline minimum problem asks us to maintain a dynamic set T of elements from the domain {1, 2, …, n} under the operations Insert and ExtractMin. We are given a sequence S of n Insert and m ExtractMin calls, where each key in {1, 2, …, n} is inserted exactly once. We wish to determine which key is returned by each ExtractMin call. Specifically, we wish to fill in an array extracted[1‥m], where for i = 1, 2, …, m, extracted[i] is the key returned by the ith ExtractMin call. The problem is “offline” in the sense that we are allowed to process the entire sequence S before determining any of the returned keys.
a. In the following instance of the offline minimum problem, each operation Insert(i) is represented by the value of i and each ExtractMin is represented by the letter E:
4, 8, E, 3, E, 9, 2, 6, E, E, E, 1, 7, E, 5.
Fill in the correct values in the extracted array.
To develop an algorithm for this problem, we break the sequence S into homogeneous subsequences. That is, we represent S by
\(\mathrm{I}_1\), E, \(\mathrm{I}_2\), E, \(\mathrm{I}_3\), …, \(\mathrm{I}_m\), E, \(\mathrm{I}_{m + 1}\);
where each E represents a single ExtractMin call and each \(\mathrm{I}_j\) represents a (possibly empty) sequence of Insert calls. For each subsequence \(\mathrm{I}_j\), we initially place the keys inserted by these operations into a set \(K_j\), which is empty if \(\mathrm{I}_j\) is empty. We then do the following:
OffLineMinimum(m, n)
 for i = 1 to n
 determine j such that i ∈ \(K_j\)
 if j ≠ m + 1
 extracted[j] = i
 let l be the smallest value greater than j for which set \(K_l\) exists
 \(K_l\) = \(K_j\) ∪ \(K_l\), destroying \(K_j\)
 return extracted
 b. Argue that the array extracted returned by OffLineMinimum is correct.
 c. Describe how to implement OffLineMinimum efficiently with a disjointset data structure. Give a tight bound on the worstcase running time of your implementation.
 a. extracted = {4, 3, 2, 6, 8, 1}.
 b. Skipped.
 c. Skipped.
212 Depth determination
In the depthdetermination problem, we maintain a forest \(\mathcal{F}\) = {\(T_i\)} of rooted trees under three operations:
MakeTree(v) creates a tree whose only node is v.
FindDepth(v) returns the depth of node v within its tree.
Graft(r, v) makes node r, which is assumed to be the root of a tree, become the child of node v, which is assumed to be in a different tree than r but may or may not itself be a root.
 a. Suppose that we use a tree representation similar to a disjointset forest: v.p is the parent of node v, except that v.p = v if v is a root. Suppose further that we implement Graft(r, v) by setting r.p = v and FindDepth(v) by following the find path up to the root, returning a count of all nodes other than v encountered. Show that the worstcase running time of a sequence of m MakeTree, FindDepth, and Graft operations is Θ(\(m^2\)).
By using the unionbyrank and pathcompression heuristics, we can reduce the worstcase running time. We use the disjointset forest \(\mathcal{S}\) = {\(S_i\)}, where each set \(S_i\) (which is itself a tree) corresponds to a tree \(T_i\) in the forest \(\mathcal{F}\). The tree structure within a set \(S_i\), however, does not necessarily correspond to that of \(T_i\). In fact, the implementation of \(S_i\) does not record the exact parentchild relationships but nevertheless allows us to determine any node’s depth in \(T_i\).
The key idea is to maintain in each node v a “pseudodistance” v.d, which is defined so that the sum of the pseudodistances along the simple path from v to the root of its set \(S_i\) equals the depth of v in \(T_i\). That is, if the simple path from v to its root in \(S_i\) is \(v_0\), \(v_1\), …, \(v_k\), where \(v_0\) = v and \(v_k\) is \(S_i\)’s root, then the depth of v in \(T_i\) is \(∑_{j = 0}^k v_j{.}d\).
 b. Give an implementation of MakeTree.
 c. Show how to modify FindSet to implement FindDepth. Your implementation should perform path compression, and its running time should be linear in the length of the find path. Make sure that your implementation updates pseudodistances correctly.
 d. Show how to implement Graft(r, v), which combines the sets containing r and v, by modifying the Union and Link procedures. Make sure that your implementation updates pseudodistances correctly. Note that the root of a set \(S_i\) is not necessarily the root of the corresponding tree \(T_i\).
 e. Give a tight bound on the worstcase running time of a sequence of m MakeTree, FindDepth, and Graft operations, n of which are MakeTree operations.
Skipped.
213 Tarjan’s offline leastcommonancestors algorithm
The least common ancestor of two nodes u and v in a rooted tree T is the node w that is an ancestor of both u and v and that has the greatest depth in T. In the offline leastcommonancestors problem, we are given a rooted tree T and an arbitrary set P = {{u, v}} of unordered pairs of nodes in T, and we wish to determine the least common ancestor of each pair in P.
To solve the offline leastcommonancestors problem, the following procedure performs a tree walk of T with the initial call LCA(T.root). We assume that each node is colored white prior to the walk.
LCA(u)
 MakeSet(u)
 FindSet(u).ancestor = u
 for each child v of u in T
 LCA(v)
 Union(u, v)
 FindSet(u).ancestor = u
 u.color = black
 for each node v such that {u, v} ∈ P
 if v.color == black
 print “The least common ancestor of” u “and” v “is” FindSet(v).ancestor
 a. Argue that line 10 executes exactly once for each pair {u, v} ∈ P.
 b. Argue that at the time of the call LCA(u), the number of sets in the disjointset data structure equals the depth of u in T.
 c. Prove that LCA correctly prints the least common ancestor of u and v for each pair {u, v} ∈ P.
 d. Analyze the running time of LCA, assuming that we use the implementation of the disjointset data structure in Section 21.3.
Skipped.
VI Graph Algorithms
22 Elementary Graph Algorithms
22.1 Representations of graphs
22.11
Given an adjacencylist representation of a directed graph, how long does it take to compute the outdegree of every vertex? How long does it take to compute the indegrees?
Both take Θ(V + E) time.
22.12
Give an adjacencylist representation for a complete binary tree on 7 vertices. Give an equivalent adjacencymatrix representation. Assume that vertices are numbered from 1 to 7 as in a binary heap.
Adjacencylist:
Vertex  Adjacency list 

1  2 3 
2  1 4 5 
3  1 6 7 
4  2 
5  2 
6  3 
7  3 
Adjacencymatrix:
1  2  3  4  5  6  7  

1  0  1  1  0  0  0  0 
2  1  0  0  1  1  0  0 
3  1  0  0  0  0  1  1 
4  0  1  0  0  0  0  0 
5  0  1  0  0  0  0  0 
6  0  0  1  0  0  0  0 
7  0  0  1  0  0  0  0 
22.13
The transpose of a directed graph G = (V, E) is the graph \(G^T\) = (V, \(E^T\)), where \(E^T\) = {(v, u) ∈ V × V : (u, v) ∈ E}. Thus, \(G^T\) is G with all its edges reversed. Describe efficient algorithms for computing \(G^T\) from G, for both the adjacencylist and adjacencymatrix representations of G. Analyze the running times of your algorithms.
Solution is implemented here.
Algorithm for adjacencylist takes O(V + E) time.
Algorithm for adjacencymatrix takes O(\(\leftV\right^2\)) time.
22.14
Given an adjacencylist representation of a multigraph G = (V, E), describe an O(V + E)time algorithm to compute the adjacencylist representation of the “equivalent” undirected graph G′ = (V, E′), where E′ consists of the edges in E with all multiple edges between two vertices replaced by a single edge and with all selfloops removed.
Solution is implemented here.
22.15
The square of a directed graph G = (V, E) is the graph \(G^2\) = (V, \(E^2\)) such that (u, v) ∈ \(E^2\) if and only if G contains a path with at most two edges between u and v. Describe efficient algorithms for computing \(G^2\) from G for both the adjacencylist and adjacencymatrix representations of G. Analyze the running times of your algorithms.
Solution is implemented here.
Not sure what running time complexity is the algorithm for adjacencylist.
The running time complexity of the algorithm for adjacencymatrix is O(\(n^3\)).
22.16
Most graph algorithms that take an adjacencymatrix representation as input require time Ω(\(V^2\)), but there are some exceptions. Show how to determine whether a directed graph G contains a universal sink—a vertex with indegree V  1 and outdegree 0—in time O(V), given an adjacency matrix for G.
Solution is implemented here.
22.17
The incidence matrix of a directed graph G = (V, E) with no selfloops is a V × E matrix B = (\(b_{ij}\)) such that
\(b_{ij} = \begin{cases} 1&\text{if edge $j$ leaves vertex $i$,}\\ 1&\text{if edge $j$ enters vertex $i$,}\\ 0&\text{otherwise.}\\ \end{cases}\)
Describe what the entries of the matrix product \(B B^T\) represent, where \(B^T\) is the transpose of B.
\((B^T)_{i i}\) represents the number of edges connected to vertex i, and \((B^T)_{i j}\) (where i ≠ j) represents the number of edges connecting edge i and edge j.
22.18
Suppose that instead of a linked list, each array entry Adj[u] is a hash table containing the vertices v for which (u, v) ∈ E. If all edge lookups are equally likely, what is the expected time to determine whether an edge is in the graph? What disadvantages does this scheme have? Suggest an alternate data structure for each edge list that solves these problems. Does your alternative have disadvantages compared to the hash table?
The expected time to determine whether an edge is in the graph is O(1).
Advantages and disadvantages are skipped.
22.2 Breadthfirst search
22.21
Show the d and π values that result from running breadthfirst search on the directed graph of Figure 22.2(a), using vertex 3 as the source.
Skipped.
22.22
Show the d and π values that result from running breadthfirst search on the undirected graph of Figure 22.3, using vertex u as the source.
Skipped.
22.23
Show that using a single bit to store each vertex color suffices by arguing that the BFS procedure would produce the same result if line 18 were removed.
A single bit can represent two state: white and gray.
The color is only checked for white at line 13, and after we remove line 18, the color of all nodes will stay gray instead of black, which does not affect the traversal process.
22.24
What is the running time of BFS if we represent its input graph by an adjacency matrix and modify the algorithm to handle this form of input?
Θ(\(V^2\)).
22.25
Argue that in a breadthfirst search, the value u.d assigned to a vertex u is independent of the order in which the vertices appear in each adjacency list. Using Figure 22.3 as an example, show that the breadthfirst tree computed by BFS can depend on the ordering within adjacency lists.
Skipped.
22.26
Give an example of a directed graph G = (V, E), a source vertex s ∈ V, and a set of tree edges \(E_\pi\) ⊆ E such that for each vertex v ∈ V, the unique simple path in the graph (V, \(E_\pi\)) from s to v is a shortest path in G, yet the set of edges \(E_\pi\) cannot be produced by running BFS on G, no matter how the vertices are ordered in each adjacency list.
Graph:
a───c
╱ ╲ ╱
s ╳
╲ ╱ ╲
b───d
Tree:
a───c
╱
s
╲
b───d
22.27
There are two types of professional wrestlers: “babyfaces” (“good guys”) and “heels” (“bad guys”). Between any pair of professional wrestlers, there may or may not be a rivalry. Suppose we have n professional wrestlers and we have a list of r pairs of wrestlers for which there are rivalries. Give an O(n + r)time algorithm that determines whether it is possible to designate some of the wrestlers as babyfaces and the remainder as heels such that each rivalry is between a babyface and a heel. If it is possible to perform such a designation, your algorithm should produce it.
Solution is implemented here.
22.28 ★
The diameter of a tree T = (V, E) is defined as \(\max_{u, v ∈ V} \delta(u, v)\), that is, the largest of all shortestpath distances in the tree. Give an efficient algorithm to compute the diameter of a tree, and analyze the running time of your algorithm.
Skipped.
22.29
Let G = (V, E) be a connected, undirected graph. Give an O(V + E)time algorithm to compute a path in G that traverses each edge in E exactly once in each direction. Describe how you can find your way out of a maze if you are given a large supply of pennies.
 Solution is implemented here.
 Skipped.
22.3 Depthfirst search
22.31
Make a 3by3 chart with row and column labels white, gray, and black. In each cell (i, j), indicate whether, at any point during a depthfirst search of a directed graph, there can be an edge from a vertex of color i to a vertex of color j. For each possible edge, indicate what edge types it can be. Make a second such chart for depthfirst search of an undirected graph.
Directed graph:
white  gray  black  

white  Tree, Back, Forward, Cross  Back, Cross  Cross 
gray  Tree, Forward  Tree, Back, Forward  Tree, Forward, Cross 
black    Back  Tree, Back, Forward, Cross 
Undirected graph:
white  gray  black  

white  Tree, Back  Tree, Back   
gray  Tree, Back  Tree, Back  Tree, Back 
black    Tree, Back  Tree, Back 
22.32
Show how depthfirst search works on the graph of Figure 22.6. Assume that the for loop of lines 5–7 of the DFS procedure considers the vertices in alphabetical order, and assume that each adjacency list is ordered alphabetically. Show the discovery and finishing times for each vertex, and show the classification of each edge.
q←──────┐ r ╱│╲ │ ╱ ╲ ╱ │ ╲ │ ╱ ╲ ↙ │ ↘ │ ╱ ↘ s │ t │ ╱ u ╱ ↖ │ │╲ │ ╱ │ ╱ ╲ │ │ ╲ │ ╱ │ ↙ ╲↓ ↓ ↘│↙ │ v──────→w x y←──────────┘ ⇵ z
Figure 22.6 A directed graph for use in Exercises 22.32 and 22.52.
Skipped.
22.33
Show the parenthesis structure of the depthfirst search of Figure 22.4.
(u (v (y (x x) y) v) u) (w (z z) w)
22.34
Show that using a single bit to store each vertex color suffices by arguing that the DFS procedure would produce the same result if line 8 of DFSVisit was removed.
This is similar to exercise 22.23.
22.35
Show that edge (u, v) is
 a. a tree edge or forward edge if and only if u.d < v.d < v.f < u.f,
 b. a back edge if and only if v.d ≤ u.d < u.f ≤ v.f, and
 c. a cross edge if and only if v.d < v.f < u.d < u.f.
Skipped.
22.36
Show that in an undirected graph, classifying an edge (u, v) as a tree edge or a back edge according to whether (u, v) or (v, u) is encountered first during the depthfirst search is equivalent to classifying it according to the ordering of the four types in the classification scheme.
Skipped.
22.37
Rewrite the procedure DFS, using a stack to eliminate recursion.
Solution is implemented here.
22.38
Give a counterexample to the conjecture that if a directed graph G contains a path from u to v, and if u.d < v.d in a depthfirst search of G, then v is a descendant of u in the depthfirst forest produced.
u ⇄ w → v
Visit from w node, we can get both u and v being w’s children.
22.39
Give a counterexample to the conjecture that if a directed graph G contains a path from u to v, then any depthfirst search must result in v.d ≤ u.f.
u ⇄ w → v
Visit from w node, then explore u before v.
22.310
Modify the pseudocode for depthfirst search so that it prints out every edge in the directed graph G, together with its type. Show what modifications, if any, you need to make if G is undirected.
Solution is implemented here.
22.311
Explain how a vertex u of a directed graph can end up in a depthfirst tree containing only u, even though u has both incoming and outgoing edges in G.
a → u → b
Discover the graph in the order b → u → a, then u is in a depthfirst tree containing only u.
22.312
Show that we can use a depthfirst search of an undirected graph G to identify the connected components of G, and that the depthfirst forest contains as many trees as G has connected components. More precisely, show how to modify depthfirst search so that it assigns to each vertex v an integer label v.cc between 1 and k, where k is the number of connected components of G, such that u.cc = v.cc if and only if u and v are in the same connected component.
Solution is implemented here.
22.313 ★
A directed graph G = (V, E) is singly connected if u ↝ v implies that G contains at most one simple path from u to v for all vertices u, v ∈ V. Give an efficient algorithm to determine whether or not a directed graph is singly connected.
I suspect that “singly connected” means that:
 There are no forward edges,
 and there are no cross edges that belongs to the same depthfirst tree.
But I can’t prove it.
22.4 Topological sort
22.41
Show the ordering of vertices produced by TopologicalSort when it is run on the dag of Figure 22.8, under the assumption of Exercise 22.32.
Skipped.
22.42
Give a lineartime algorithm that takes as input a directed acyclic graph G = (V, E) and two vertices s and t, and returns the number of simple paths from s to t in G. For example, the directed acyclic graph of Figure 22.8 contains exactly four simple paths from vertex p to vertex v: p o v, p o r y v, p o s r y v, and p s r y v. (Your algorithm needs only to count the simple paths, not list them.)
Solution is implemented here.
22.43
Give an algorithm that determines whether or not a given undirected graph G = (V, E) contains a simple cycle. Your algorithm should run in O(V) time, independent of E.
Solution is implemented here.
22.44
Prove or disprove: If a directed graph G contains cycles, then TopologicalSort(G) produces a vertex ordering that minimizes the number of “bad” edges that are inconsistent with the ordering produced.
Skipped.
22.45
Another way to perform topological sorting on a directed acyclic graph G = (V, E) is to repeatedly find a vertex of indegree 0, output it, and remove it and all of its outgoing edges from the graph. Explain how to implement this idea so that it runs in time O(V + E). What happens to this algorithm if G has cycles?
Solution is implemented here.
22.5 Strongly connected components
22.51
How can the number of strongly connected components of a graph change if a new edge is added?
Skipped.
22.52
Show how the procedure StronglyConnectedComponents works on the graph of Figure 22.6. Specifically, show the finishing times computed in line 1 and the forest produced in line 3. Assume that the loop of lines 5–7 of DFS considers vertices in alphabetical order and that the adjacency lists are in alphabetical order.
Skipped.
22.53
Professor Bacon claims that the algorithm for strongly connected components would be simpler if it used the original (instead of the transpose) graph in the second depthfirst search and scanned the vertices in order of increasing finishing times. Does this simpler algorithm always produce correct results?
No.
For this graph:
u ⇄ w → v
If the first DFS starts from node w, and visit u first, then u is the first finished node. The second DFS then starts from u, which will visit w and v, so this algorithm will consider u, v and w as a single strongly connected component, which is wrong.
22.54
Prove that for any directed graph G, we have \(((G^T)^{SCC})^T = G^{SCC}\). That is, the transpose of the component graph of \(G^T\) is the same as the component graph of G.
Skipped.
22.55
Give an O(V + E)time algorithm to compute the component graph of a directed graph G = (V, E). Make sure that there is at most one edge between two vertices in the component graph your algorithm produces.
Solution is implemented here;
22.56
Given a directed graph G = (V, E), explain how to create another graph G′ = (V, E′) such that (a) G′ has the same strongly connected components as G, (b) G′ has the same component graph as G, and (c) E′ is as small as possible. Describe a fast algorithm to compute G′.
Skipped.
22.57
A directed graph G = (V, E) is semiconnected if, for all pairs of vertices u, v ∈ V, we have u ↝ v or v ↝ u. Give an efficient algorithm to determine whether or not G is semiconnected. Prove that your algorithm is correct, and analyze its running time.
Skipped.
22.X Problems
221 classifying edges by breadthfirst search
A depthfirst forest classifies the edges of a graph into tree, back, forward, and cross edges. A breadthfirst tree can also be used to classify the edges reachable from the source of the search into the same four categories.
a. Prove that in a breadthfirst search of an undirected graph, the following properties hold:
 There are no back edges and no forward edges.
 For each tree edge (u, v), we have v.d = u.d + 1.
 For each cross edge (u, v), we have v.d = u.d or v.d = u.d + 1.
b. Prove that in a breadthfirst search of a directed graph, the following properties hold:
 There are no forward edges.
 For each tree edge (u, v), we have v.d = u.d + 1.
 For each cross edge (u, v), we have v.d ≤ u.d + 1.
 For each back edge (u, v), we have 0 ≤ v.d ≤ u.d.
Skipped.
222 Articulation points, bridges, and biconnected components
Let G = (V, E) be a connected, undirected graph. An articulation point of G is a vertex whose removal disconnects G. A bridge of G is an edge whose removal disconnects G. A biconnected component of G is a maximal set of edges such that any two edges in the set lie on a common simple cycle. Figure 22.10 illustrates these definitions. We can determine articulation points, bridges, and biconnected components using depthfirst search. Let \(G_\pi\) = (V, \(E_\pi\)) be a depthfirst tree of G.
1 5─6 11 19──────┐ ╱│╲ ╲│ ╱ ╲ │╲ │ 0─┼─3───4───9 13────14─┼─16───20 ╲│╱ ╱│ ╲ ╱ ╲│╱ 2 7─8 12 21 15 │ │ ╲ ╱ ╲ 10 22─17 18
Figure 22.10 The articulation points, bridges, and biconnected components of a connected, undirected graph for use in Problem 222. The articulation points are the heavily shaded vertices, the bridges are the heavily shaded edges, and the biconnected components are the edges in the shaded regions, with a bcc numbering shown.
a. Prove that the root of \(G_\pi\) is an articulation point of G if and only if it has at least two children in \(G_\pi\).
b. Let v be a nonroot vertex of \(G_\pi\). Prove that v is an articulation point of G if and only if v has a child s such that there is no back edge from s or any descendant of s to a proper ancestor of v.
c. Let
v.low = \(\min \begin{cases}v{.}d\text{,}\\ w{.}d : \text{$(u, w)$ is a back edge for some descendant $u$ of $v$}\end{cases}\)
Show how to compute v.low for all vertices v ∈ V in O(E) time.
d. Show how to compute all articulation points in O(E) time.
e. Prove that an edge of G is a bridge if and only if it does not lie on any simple cycle of G.
f. Show how to compute all the bridges of G in O(E) time.
g. Prove that the biconnected components of G partition the nonbridge edges of G.
h. Give an O(E)time algorithm to label each edge e of G with a positive integer e.bcc such that e.bcc = e′.bcc if and only if e and e′ are in the same biconnected component.
 a.
 Sufficient condition: If the root has at least two children, then vertices that are descendant of one child can only reach vertices that are descendant of other children through the root. If we remove the root, the reachability will be broken, so the root is an articulation point.
 Necessary condition: Proof by contradiction: Suppose the root has less than two children:
 If the root has no child, then the graph only has one vertex, so the root is not an articulation point.
 If the root has one child, then all vertices except the root is descendant of the child, so if we remote the root, the reachability will be kept, so the root is not an articulation point.
 b.
 Sufficient condition: If the vertex v has a child s such that there is no back edge from s or any descendant of s to a proper ancestor of v, then for descendants of s can only reach v’s ancestors through v, after removing v, the reachability will be broken.
 Necessary condition: Proof by contradiction: Suppose all children of the vertex v has back edges from their descendants to some v’s ancestors, then all descendants of v’s children can reach the root without going through v. After removing v, all the other vertices can still reach root, so the reachability is kept.
 c. Solution is implemented here.
 d. Solution is implemented here.
 e. Skipped.
 f. Solution is implemented here.
 g. Skipped.
 h. Solution is implemented here.
223 Euler tour
An Euler tour of a strongly connected, directed graph G = (V, E) is a cycle that traverses each edge of G exactly once, although it may visit a vertex more than once.
 a. Show that G has an Euler tour if and only if indegree(v) = outdegree(v) for each vertex v ∈ V.
 b. Describe an O(E)time algorithm to find an Euler tour of G if one exists. (Hint: Merge edgedisjoint cycles.)
 a. Skipped.
 b. Solution is implemented here, but I am not 100% sure that it is correct.
224 Reachability
Let G = (V, E) be a directed graph in which each vertex u ∈ V is labeled with a unique integer L(u) from the set {1, 2, …, V}. For each vertex u ∈ V, let R(u) = {v ∈ V : u ↝ v} be the set of vertices that are reachable from u. Define min(u) to be the vertex in R(u) whose label is minimum, i.e., min(u) is the vertex v such that L(v) = min{L(w) : w ∈ R(u)}. Give an O(V + E)time algorithm that computes min(u) for all vertices u ∈ V.
Solution is implemented here.
23 Minimum Spanning Trees
23.1 Growing a minimum spanning tree
23.11
Let (u, v) be a minimumweight edge in a connected graph G. Show that (u, v) belongs to some minimum spanning tree of G.
Proof by contradiction:
Suppose (u, v) does not belongs to any minimum spanning tree of G, for any minimum spanning tree of G, there must be a path in a tree from u to v. Let p be the vertex after u in the path. If we replace (u, p) with (u, v) in the tree, we can get a new tree with weight less than or equal to the original tree. So the new tree is a minimum spanning tree, which contradicts our supposition, so (u, v) must belong to some minimum spanning tree of G.
23.12
Professor Sabatier conjectures the following converse of Theorem 23.1. Let G = (V, E) be a connected, undirected graph with a realvalued weight function w defined on E. Let A be a subset of E that is included in some minimum spanning tree for G, let (S, V  S) be any cut of G that respects A, and let (u, v) be a safe edge for A crossing (S, V  S). Then, (u, v) is a light edge for the cut. Show that the professor’s conjecture is incorrect by giving a counterexample.
Example graph:
2
a─────b
│
1│
│ 1
c─────d
Let A = {(a, c)}, S = {a, c}, then (a, b) is a safe edge, but it is not a light edge.
23.13
Show that if an edge (u, v) is contained in some minimum spanning tree, then it is a light edge crossing some cut of the graph.
Proof by contradiction:
Suppose there is a minimum spanning tree that contains an edge (u, v) that is not a light edge crossing any cuts, then (u, v) will separate the minimum spanning tree into two subtrees. Since (u, v) is not a light edge, there will be a light edge cutting the two subtrees. Replace (u, v) with the light edge in the minimum spanning tree, we can get a new tree with less weight. This contradicts our supposition of the minimum spanning tree. So the claim holds.
23.14
Give a simple example of a connected graph such that the set of edges {(u, v) : there exists a cut (S, V  S) such that (u, v) is a light edge crossing (S, V  S)} does not form a minimum spanning tree.
1
a─────b
 ╱
 ╱
1 ╱1
 ╱
╱
c
23.15
Let e be a maximumweight edge on some cycle of connected graph G = (V, E). Prove that there is a minimum spanning tree of G′ = (V, E  {e}) that is also a minimum spanning tree of G. That is, there is a minimum spanning tree of G that does not include e.
Proof by contradiction: Suppose that all minimum spanning trees of G include e, then for any minimum spanning tree, removing e will break the spanning tree into two subtrees. Then, there must be an edge x on the cycle other than e connecting the two subtrees. Adding x to the two subtrees, we get a new tree with weight less than or equal to the original minimal spanning tree, which means we get a new minimal spanning tree that does not contain e. This contradicts our supposition, so the claim holds.
23.16
Show that a graph has a unique minimum spanning tree if, for every cut of the graph, there is a unique light edge crossing the cut. Show that the converse is not true by giving a counterexample.
Theorem: For any minimum spanning tree T of a graph G = (V, E), for any cut of G, edges in T cross the cut must contain a light edge. Proof: For any cut (S, V  S), let (u, v) be a light edge cross the cut, then there is a path in T that connects u and v, and there is an edge e on the path crosses the cut. If e is not a light edge, we can replace it with the light edge without breaking the connectivity to get a tree with lesser weight. So there must be an light edge cross any cuts of the tree in T.

Let T and U be any two minimum spanning trees, for any edge (u, v) in T, removing it will break T into two subtrees, and according to exercise 23.13, we know that (u, v) must be a light edge cross the cut of two subtrees. Then since edges cross the same cut in U must contain a light edge, we know that (u, v) must be in U. So any edge in T is in U, T and U must be the same tree, so the minimum spanning tree is unique.

See the following graph:
1 1 a─────b─────c
The cut ({a, c}, {b}) has two light edges, but the minimum spanning tree is still unique.
23.17
Argue that if all edge weights of a graph are positive, then any subset of edges that connects all vertices and has minimum total weight must be a tree. Give an example to show that the same conclusion does not follow if we allow some weights to be nonpositive.

Proof by contradiction: Suppose the subset is not a tree, there must be cycles in the subset. Since the weights are positive, remove any edge from the cycle will not break the connection but we will get a subset with lesser weight, this contradicts out supposition.

See the following graph:
1 a─────b  ╱  ╱ 1 ╱1  ╱ ╱ c
23.18
Let T be a minimum spanning tree of a graph G, and let L be the sorted list of the edge weights of T. Show that for any other minimum spanning tree T′ of G, the list L is also the sorted list of edge weights of T′.
For any two different minimum spanning trees T and T′, there must be some edge (u, v) exist in T, but in T′. (u, v) in T partitions the graph into a cut. Since T′ does not contain edge (u, v), there must be a path in T′ crossing the cut that connects u and v, thus the must be an edge (x, y) in T′ crossing the cut.
Now we show that the weight of (x, y) must equal to the weight of (u, v). Removing edge (x, y) will breaking T′ into two subtrees, each containing u and v. We can then connect these two subtree into a new tree by add the edge (u, v). Since T′ is a minimum spanning tree, we can not get a new tree with less weight, so the weight of (u, v) must be greater than or equal to the weight of (x, y). On the other hand, we can replace (u, v) with (x, y) in T to get a new tree, since T is a minimum spanning tree, we know that the weight of (x, y) must be greater than or equal to the weight of (u, v). So the weight of (x, y) must equal to the weight of (u, v).
The procedure above can transform T into another tree by replacing an edge with another edge of same weight. The new tree has the same sorted edge weight list as T but with one less different edge with T′. We can apply the procedure repeatedly until we transform T into T′. The whole procedure does not change the edge sorted weight list, which means T and T′ have the same sorted edge weight list.
23.19
Let T be a minimum spanning tree of a graph G = (V, E), and let V′ be a subset of V. Let T′ be the subgraph of T induced by V′, and let G′ be the subgraph of G induced by V′. Show that if T′ is connected, then T′ is a minimum spanning tree of G′.
Proof by contradiction:
If T′ is not a minimum spanning tree of G′, there must be a minimum spanning tree U of G′ with weight less than T′. We can replace T′ in T with U to get a tree with lesser weight. This is impossible since T is already a minimum spanning tree, so T′ must be a minimum spanning tree of G′.
23.110
Given a graph G and a minimum spanning tree T, suppose that we decrease the weight of one of the edges in T. Show that T is still a minimum spanning tree for G. More formally, let T be a minimum spanning tree for G with edge weights given by weight function w. Choose one edge (x, y) ∈ T and a positive number k, and define the weight function w′ by
w′ = \(\begin{cases} w(u, v) & \text{if $(u, v) ≠ (x, y)$,}\\ w(x, y)  k & \text{if $(u, v) = (x, y)$.} \end{cases}\)
Show that T is a minimum spanning tree for G with edge weights given by w′.
After decreasing the weight of one edge by k, the weights of all spanning trees can decrease at most by k, depdending on whether the spanning tree contains the weight decreased edge. So the weight of the minimal spanning tree can also decrease at most by k. We also know that the weight of the original minimal spanning tree decreases exactly by k, so it must be a minimal spanning tree in the new graph, since we just showed spanning trees can not decrease by more than k.
23.111 ★
Given a graph G and a minimum spanning tree T, suppose that we decrease the weight of one of the edges not in T. Give an algorithm for finding the minimum spanning tree in the modified graph.
Let the edge that has weight decreased be (u, v). First find a path in T that connects u and v. Let (x, y) be some maximal weighted edge on the path.
 If the weight of (u, v) after decreasing is less than the weight of (x, y), replace (x, y) with (u, v), we can get the new minimal spanning tree.
 Otherwise, the minimal spanning tree remains the same.
23.2 The algorithms of Kruskal and Prim
23.21
Kruskal’s algorithm can return different spanning trees for the same input graph G, depending on how it breaks ties when the edges are sorted into order. Show that for each minimum spanning tree T of G, there is a way to sort the edges of G in Kruskal’s algorithm so that the algorithm returns T.
Make sure that in each chunk of edges of the same weight, edges in T comes before edges not in T.
23.22
Suppose that we represent the graph G = (V, E) as an adjacency matrix. Give a simple implementation of Prim’s algorithm for this case that runs in O(\(V^2\)) time.
Solution is implemented here.
23.23
For a sparse graph G = (V, E), where E = Θ(V), is the implementation of Prim’s algorithm with a Fibonacci heap asymptotically faster than the binaryheap implementation? What about for a dense graph, where E = (\(V^2\))? How must the sizes E and V be related for the Fibonacciheap implementation to be asymptotically faster than the binaryheap implementation?
Skipped.
23.24
Suppose that all edge weights in a graph are integers in the range from 1 to V. How fast can you make Kruskal’s algorithm run? What if the edge weights are integers in the range from 1 to W for some constant W?
Skipped.
23.25
Suppose that all edge weights in a graph are integers in the range from 1 to V. How fast can you make Prim’s algorithm run? What if the edge weights are integers in the range from 1 to W for some constant W?
Skipped.
23.26 ★
Suppose that the edge weights in a graph are uniformly distributed over the halfopen interval [0, 1). Which algorithm, Kruskal’s or Prim’s, can you make run faster?
Skipped.
23.27 ★
Suppose that a graph G has a minimum spanning tree already computed. How quickly can we update the minimum spanning tree if we add a new vertex and incident edges to G?
Skipped.
23.28
Professor Borden proposes a new divideandconquer algorithm for computing minimum spanning trees, which goes as follows. Given a graph G = (V, E), partition the set V of vertices into two sets \(V_1\) and \(V_2\) such that \(V_1\) and \(V_2\) differ by at most 1. Let \(E_1\) be the set of edges that are incident only on vertices in \(V_1\), and let \(E_2\) be the set of edges that are incident only on vertices in \(V_2\). Recursively solve a minimumspanningtree problem on each of the two subgraphs \(G_1\) = (\(V_1\), \(E_1\)) and \(G_2\) = (\(V_2\), \(E_2\)). Finally, select the minimumweight edge in E that crosses the cut (\(V_1\), \(V_2\)), and use this edge to unite the resulting two minimum spanning trees into a single spanning tree.
Either argue that the algorithm correctly computes a minimum spanning tree of G, or provide an example for which the algorithm fails.
This algorithms fails to compute a minimum spanning tree of the following graph:
1
a─────b
│ │
1│ │1
│ │
c─────d
2
If we partition the graph above into a subgraph containing {a, b} and a subgraph containing {c, d}, the result will be:
1
a─────b
│
1│
│
c─────d
2
But the correct answer should be:
1
a─────b
│ │
1│ │1
│ │
c d
23.X Problems
231 Secondbest minimum spanning tree
Let G = (V, E) be an undirected, connected graph whose weight function is w: E → ℝ, and suppose that E ≥ V and all edge weights are distinct.
We define a secondbest minimum spanning tree as follows. Let 𝒯 be the set of all spanning trees of G, and let T′ be a minimum spanning tree of G. Then a secondbest minimum spanning tree is a spanning tree T such that w(T) = \(\min_{T'' ∈ 𝒯  \{T'\}} \{w(T'')\}\).
 a. Show that the minimum spanning tree is unique, but that the secondbest minimum spanning tree need not be unique.
 b. Let T be the minimum spanning tree of G. Prove that G contains edges (u, v) ∈ T and (x, y) ∉ T such that T  {(u, v)} ∪ {(x, y)} is a secondbest minimum spanning tree of G.
 c. Let T be a spanning tree of G and, for any two vertices u, v ∈ V, let max[u, v] denote an edge of maximum weight on the unique simple path between u and v in T. Describe an O(\(V^2\))time algorithm that, given T, computes max[u, v] for all u, v ∈ V.
 d. Give an efficient algorithm to compute the secondbest minimum spanning tree of G.
Skipped.
232 Minimum spanning tree in sparse graphs
For a very sparse connected graph G = (V, E), we can further improve upon the O(E + V lg V) running time of Prim’s algorithm with Fibonacci heaps by preprocessing G to decrease the number of vertices before running Prim’s algorithm. In particular, we choose, for each vertex u, the minimumweight edge (u, v) incident on u, and we put (u, v) into the minimum spanning tree under construction. We then contract all chosen edges (see Section B.4). Rather than contracting these edges one at a time, we first identify sets of vertices that are united into the same new vertex. Then we create the graph that would have resulted from contracting these edges one at a time, but we do so by “renaming” edges according to the sets into which their endpoints were placed. Several edges from the original graph may be renamed the same as each other. In such a case, only one edge results, and its weight is the minimum of the weights of the corresponding original edges.
Initially, we set the minimum spanning tree T being constructed to be empty, and for each edge (u, v) ∈ E, we initialize the attributes (u, v).orig = (u, v) and (u, v).c = w(u, v). We use the orig attribute to reference the edge from the initial graph that is associated with an edge in the contracted graph. The c attribute holds the weight of an edge, and as edges are contracted, we update it according to the above scheme for choosing edge weights. The procedure MSTReduce takes inputs G and T , and it returns a contracted graph G′ with updated attributes orig′ and c′. The procedure also accumulates edges of G into the minimum spanning tree T.
MSTReduce(G, T)
 for each v ∈ G.V
 v.mark = false
 MakeSet(v)
 for each u ∈ G.V
 if u.mark == false
 choose v ∈ G.Adj[u] such that (u, v).c is minimized
 Union(u, v)
 T = T ∪ {(u, v).orig}
 u.mark = v.mark = true
 G′.V = {FindSet(v): v ∈ G.V}
 G′.E = ∅
 for each (x, y) ∈ G.E
 u = FindSet(x)
 v = FindSet(y)
 if u ≠ v
 if (u, v) ∉ G′.E
 G′.E = G′.E ∪ {(u, v)}
 (u, v).orig′ = (x, y).orig
 (u, v).c′ = (x, y).c
 else if (x, y).c < (u, v).c′
 (u, v).orig′ = (x, y).orig
 (u, v).c′ = (x, y).c
 construct adjacency lists G′.Adj for G′
 return G′ and T
 a. Let T be the set of edges returned by MSTReduce, and let A be the minimum spanning tree of the graph G′ formed by the call MSTPrim(G′, c′, r), where c′ is the weight attribute on the edges of G′.E and r is any vertex in G′.V. Prove that T ∪ {(x, y).orig′: (x, y) ∈ A} is a minimum spanning tree of G.
 b. Argue that G′.V ≤ V / 2.
 c. Show how to implement MSTReduce so that it runs in O(E) time. (Hint: Use simple data structures.)
 d. Suppose that we run k phases of MSTReduce, using the output G′ produced by one phase as the input G to the next phase and accumulating edges in T. Argue that the overall running time of the k phases is O(k E).
 e. Suppose that after running k phases of MSTReduce, as in part (d), we run Prim’s algorithm by calling MSTPrim(G′, c′, r), where G′, with weight attribute c′, is returned by the last phase and r is any vertex in G′.V. Show how to pick k so that the overall running time is O(E lg lg V). Argue that your choice of k minimizes the overall asymptotic running time.
 f. For what values of E (in terms of V) does Prim’s algorithm with preprocessing asymptotically beat Prim’s algorithm without preprocessing?
Skipped.
233 Bottleneck spanning tree
A bottleneck spanning tree T of an undirected graph G is a spanning tree of G whose largest edge weight is minimum over all spanning trees of G. We say that the value of the bottleneck spanning tree is the weight of the maximumweight edge in T.
 a. Argue that a minimum spanning tree is a bottleneck spanning tree.
Part (a) shows that finding a bottleneck spanning tree is no harder than finding a minimum spanning tree. In the remaining parts, we will show how to find a bottleneck spanning tree in linear time.
 b. Give a lineartime algorithm that given a graph G and an integer b, determines whether the value of the bottleneck spanning tree is at most b.
 c. Use your algorithm for part (b) as a subroutine in a lineartime algorithm for the bottleneckspanningtree problem. (Hint: You may want to use a subroutine that contracts sets of edges, as in the MSTReduce procedure described in Problem 232.)
 a. Proof by contradiction. Suppose there exists a spanning tree T whose largest edge is less than that of a minimum spanning tree T′, we can replace T′’s largest edge with some edge from T to get a smaller spanning tree, which contradicts the supposition that T′ is a minimum spanning tree, so all spanning tree’s largest edge must be greater than or equal to that of T′, which means T′ is a bottleneck spanning tree.
 b. Skipped.
 c. Skipped.
234 Alternative minimumspanningtree algorithms
In this problem, we give pseudocode for three different algorithms. Each one takes a connected graph and a weight function as input and returns a set of edges T. For each algorithm, either prove that T is a minimum spanning tree or prove that T is not necessarily a minimum spanning tree. Also describe the most efficient implementation of each algorithm, whether or not it computes a minimum spanning tree.
a. MaybeMSTA(G, w)
 sort the edges into nonincreasing order of edge weights w
 T = E
 for each edge e, taken in nonincreasing order by weight
 if T  {e} is a connected graph
 T = T  {e}
 return T
b. MaybeMSTB(G, w)
 T = ∅
 for each edge e, taken in arbitrary order
 if T ∪ {e} has no cycles
 T = T ∪ {e}
 return T
c. MaybeMSTC(G, w)
 T = ∅
 for each edge e, taken in arbitrary order
 T = T ∪ {e}
 if T has a cycle c
 let e′ be a maximumweight edge on c
 T = T  {e′}
 return T
Skipped.
24 SingleSource Shortest Paths
24.1 The BellmanFord algorithm
24.11
Run the BellmanFord algorithm on the directed graph of Figure 24.4, using vertex z as the source. In each pass, relax edges in the same order as in the figure, and show the d and π values after each pass. Now, change the weight of edge (z, x) to 4 and run the algorithm again, using s as the source.
Skipped.
24.12
Prove Corollary 24.3.
Skipped.
24.13
Given a weighted, directed graph G = (V, E) with no negativeweight cycles, let m be the maximum over all vertices v ∈ V of the minimum number of edges in a shortest path from the source s to v. (Here, the shortest path is by weight, not the number of edges.) Suggest a simple change to the BellmanFord algorithm that allows it to terminate in m + 1 passes, even if m is not known in advance.
If one pass does not update any d value, we know any succeeding passes will has no effect on the result, so we can just stop after the first pass that does not update any d value.
24.14
Modify the BellmanFord algorithm so that it sets v.d to ∞ for all vertices v for which there is a negativeweight cycle on some path from the source to v.
Solution is implemented here.
24.15 ★
Let G = (V, E) be a weighted, directed graph with weight function w: E → ℝ. Give an O(V E)time algorithm to find, for each vertex v ∈ V, the value \(δ^*(v) = \min_{u ∈ V} \{δ(u, v)\}\).
The trick is to treat v.d as the upper bound of the shortestpath weight from any u ∈ V instead of a single fixed source vertex.
Solution is implemented here.
24.16 ★
Suppose that a weighted, directed graph G = (V, E) has a negativeweight cycle. Give an efficient algorithm to list the vertices of one such cycle. Prove that your algorithm is correct.
Skipped.
24.2 Singlesource shortest paths in directed acyclic graphs
24.21
Run DagShortestPaths on the directed graph of Figure 24.5, using vertex r as the source.
Skipped.
24.22
Suppose we change line 3 of DagShortestPaths to read
3 for the first V  1 vertices, taken in topologically sorted order
Show that the procedure would remain correct.
The last vertex of a topologically sorted vertices in a DAG must has no adjacent vertices, so the last iteration does nothing, so it is safe to ignore the last pass.
24.23
The PERT chart formulation given above is somewhat unnatural. In a more natural structure, vertices would represent jobs and edges would represent sequencing constraints; that is, edge (u, v) would indicate that job u must be performed before job v. We would then assign weights to vertices, not edges. Modify the DagShortestPaths procedure so that it finds a longest path in a directed acyclic graph with weighted vertices in linear time.
Solution is implemented here.
24.24
Give an efficient algorithm to count the total number of paths in a directed acyclic graph. Analyze your algorithm.
Solution is implemented here.
24.3 Dijkstra’s algorithm
24.31
Run Dijkstra’s algorithm on the directed graph of Figure 24.2, first using vertex s as the source and then using vertex z as the source. In the style of Figure 24.6, show the d and π values and the vertices in set S after each iteration of the while loop.
Skipped.
24.32
Give a simple example of a directed graph with negativeweight edges for which Dijkstra’s algorithm produces incorrect answers. Why doesn’t the proof of Theorem 24.6 go through when negativeweight edges are allowed?
Skipped.
24.33
Suppose we change line 4 of Dijkstra’s algorithm to the following.
4 while Q > 1
This change causes the while loop to execute V  1 times instead of V times. Is this proposed algorithm correct?
Yes. Because once the queue contains exactly one vertex, we know that vertex would be extracted next, which means it will have the correct distance, which means all vertices will have correct distances, the algorithm does not need to go further.
24.34
Professor Gaedel has written a program that he claims implements Dijkstra’s algorithm. The program produces v.d and v.π for each vertex v ∈ V. Give an O(V + E)time algorithm to check the output of the professor’s program. It should determine whether the d and π attributes match those of some shortestpaths tree. You may assume that all edge weights are nonnegative.
Solution is implemented here.
24.35
Professor Newman thinks that he has worked out a simpler proof of correctness for Dijkstra’s algorithm. He claims that Dijkstra’s algorithm relaxes the edges of every shortest path in the graph in the order in which they appear on the path, and therefore the pathrelaxation property applies to every vertex reachable from the source. Show that the professor is mistaken by constructing a directed graph for which Dijkstra’s algorithm could relax the edges of a shortest path out of order.
Skipped.
24.36
We are given a directed graph G = (V, E) on which each edge (u, v) ∈ E has an associated value r(u, v), which is a real number in the range 0 ≤ r(u, v) ≤ 1 that represents the reliability of a communication channel from vertex u to vertex v. We interpret r(u, v) as the probability that the channel from u to v will not fail, and we assume that these probabilities are independent. Give an efficient algorithm to find the most reliable path between two given vertices.
Run the Dijkstra algorithm on the graph, but with edge weight of (u, v) set to log(r(u, v)).
24.37
Let G = (V, E) be a weighted, directed graph with positive weight function w: E → {1, 2, …, W} for some positive integer W, and assume that no two vertices have the same shortestpath weights from source vertex s. Now suppose that we define an unweighted, directed graph G′ = (V ∪ V′, E′) by replacing each edge (u, v) ∈ E with w(u, v) unitweight edges in series. How many vertices does G′ have? Now suppose that we run a breadthfirst search on G′. Show that the order in which the breadthfirst search of G′ colors vertices in V black is the same as the order in which Dijkstra’s algorithm extracts the vertices of V from the priority queue when it runs on G.
 G′ will have V + \(∑_{(u, v) ∈ E} w(u, v)\)  E vertices.
 Skipped.
24.38
Let G = (V, E) be a weighted, directed graph with nonnegative weight function w: E → {0, 1, …, W} for some nonnegative integer W. Modify Dijkstra’s algorithm to compute the shortest paths from a given source vertex s in O(W V + E) time.
Skipped.
24.39
Modify your algorithm from Exercise 24.38 to run in O((V + E) lg W) time. (Hint: How many distinct shortestpath estimates can there be in V  S at any point in time?)
Skipped.
24.310
Suppose that we are given a weighted, directed graph G = (V, E) in which edges that leave the source vertex s may have negative weights, all other edge weights are nonnegative, and there are no negativeweight cycles. Argue that Dijkstra’s algorithm correctly finds shortest paths from s in this graph.
Skipped.
24.4 Difference constraints and shortest paths
24.41
Find a feasible solution or determine that no feasible solution exists for the following system of difference constraints:
 \(x_1\)  \(x_2\) ≤ 1,
 \(x_1\)  \(x_4\) ≤ 4,
 \(x_2\)  \(x_3\) ≤ 2,
 \(x_2\)  \(x_5\) ≤ 7,
 \(x_2\)  \(x_6\) ≤ 5,
 \(x_3\)  \(x_6\) ≤ 10,
 \(x_4\)  \(x_2\) ≤ 2,
 \(x_5\)  \(x_1\) ≤ 1,
 \(x_5\)  \(x_4\) ≤ 3,
 \(x_6\)  \(x_3\) ≤ 8.
Skipped.
24.42
Find a feasible solution or determine that no feasible solution exists for the following system of difference constraints:
 \(x_1\)  \(x_2\) ≤ 4,
 \(x_1\)  \(x_5\) ≤ 5,
 \(x_2\)  \(x_4\) ≤ 6,
 \(x_3\)  \(x_2\) ≤ 1,
 \(x_4\)  \(x_1\) ≤ 3,
 \(x_4\)  \(x_3\) ≤ 5,
 \(x_4\)  \(x_5\) ≤ 10,
 \(x_5\)  \(x_3\) ≤ 4,
 \(x_5\)  \(x_4\) ≤ 8.
Skipped.
24.43
Can any shortestpath weight from the new vertex \(v_0\) in a constraint graph be positive? Explain.
No. Since the \(v_0\) has a zero weighted edge to every other vertices, so the shortest path weight is at most zero.
24.44
Express the singlepair shortestpath problem as a linear program.
Assume the source vertex is s, for each edge (\(v_i\), \(v_j\)), there is a constraint of \(x_j\)  \(x_i\) ≤ w(\(v_i\), \(v_j\)).
24.45
Show how to modify the BellmanFord algorithm slightly so that when we use it to solve a system of difference constraints with m inequalities on n unknowns, the running time is O(n m).
First, set the d attribute of all vertices to zero, then we can skipping relaxing edges coming from \(v_0\), because we know:
 The weight of shortest path from \(v_0\) to \(v_0\) is always zero, since there is no incoming edges going into \(v_0\).
 The weight of edges coming from \(v_0\) is zero.
So relaxing edges coming from \(v_0\) does nothing, thus we can skipping relaxing them.
24.46
Suppose that in addition to a system of difference constraints, we want to handle equality constraints of the form \(x_i\) = \(x_j\) + \(b_k\). Show how to adapt the BellmanFord algorithm to solve this variety of constraint system.
\(x_i\) = \(x_j\) + \(b_k\) is equivlent to these two constraints:
 \(x_i\)  \(x_j\) ≤ \(b_k\)
 \(x_j\)  \(x_i\) ≤ \(b_k\)
So a equality constraints problem can be converted into a difference constraints problem. Solve the converted problem, we can get the solution the equality constraints problem.
24.47
Show how to solve a system of difference constraints by a BellmanFordlike algorithm that runs on a constraint graph without the extra vertex \(v_0\).
Initialize the d attribute of each vertex, then run the BellmanFord algorithm.
List of common symbols:
×̲ΓΔΘΣΦΩαβγδεζπσωϕϵ–—’“”‥…′ℋℕℝℤℱ←↑→↓↖↘↙↝⇄⇒⇔⇣⇵∀∃∅∈∉∏∑∖∞∧∨∩∪≠≤≥⊆⊗⋂⋅⋯⌈⌉⌊⌋─│┊┌┐└┘├┤┬┴┼╱╲╳▌▬★⟨⟩𝒯
VIII Appendix: Mathematical Background
C Counting and Probability
C.2 Probability
C.21
Professor Rosencrantz flips a fair coin once. Professor Guildenstern flips a fair coin twice. What is the probability that Professor Rosencrantz obtains more heads than Professor Guildenstern?
S = {(H, HH), (H, HT), (H, TH), (H, TT), (T, HH), (T, HT), (T, TH), (T, TT)}.
A = {(H, TT)}.
Pr{A} = 1 / 8.
C.22
Prove Boole’s inequality: For any finite or countably infinite sequence of events \(A_1\), \(A_2\), …,
Pr {\(A_1\) ∪ \(A_2\) ∪ ⋯} ≤ Pr{\(A_1\)} + Pr{\(A_2\)} + ⋯. (C.19)
Proof by induction:
Base case: Pr{\(A_1\)} ≤ Pr{\(A_1\)}.
Inductive case:
\(\Pr\left\lbrace\left(\bigcup_{i = 1}^j A_i\right) ∪ A_{i + 1}\right\rbrace\)
≤ \(\Pr\left\lbrace\bigcup_{i = 1}^j A_i\right\rbrace + \Pr\left\lbrace A_{i + 1}\right\rbrace\)
≤ \(∑_{i = 1}^j \Pr\left\lbrace A_i\right\rbrace + \Pr\left\lbrace A_{i + 1}\right\rbrace\) (By induction)
= \(∑_{i = 1}^{j + 1} \Pr\left\lbrace A_i\right\rbrace\).
C.23
Suppose we shuffle a deck of 10 cards, each bearing a distinct number from 1 to 10, to mix the cards thoroughly. We then remove three cards, one at a time, from the deck. What is the probability that we select the three cards in sorted (increasing) order?
The order of the selected cards is independent of the number of cards, so the probability is 1 / 3! = 1 / 6.
C.24
Prove that Pr{A  B} + Pr{\(\bar{A}\)  B} = 1.
Pr{A  B} + Pr{\(\bar{A}\)  B}
= Pr{A ∩ B} / Pr{B} + Pr{\(\bar{A}\) ∩ B} / Pr{B}
= (Pr{A ∩ B} + Pr{\(\bar{A}\) ∩ B}) / Pr{B}
= Pr{(A ∩ B) ∪ (\(\bar{A}\) ∩ B)} / Pr{B} (Since A and \(\bar{A}\) are mutually exclusive)
= Pr{B} / Pr{B}
= 1.
C.25
Prove that for any collection of events \(A_1\), \(A_2\), …, \(A_n\),
\(\begin{aligned} \Pr\left\lbrace A_1 \cap A_2 \cap ⋯ \cap A_n\right\rbrace = &\Pr\left\lbrace A_1\right\rbrace ⋅ \Pr\left\lbrace A_2 \middle A_1\right\rbrace ⋅ \Pr\left\lbrace A_3 \middle A_1 \cap A_2\right\rbrace ⋯\\ &\Pr\left\lbrace A_n \middle A_1 \cap A_2 \cap ⋯ \cap A_{n  1}\right\rbrace\text{.} \end{aligned}\)
Proof by induction:
Base case: \(\Pr\left\lbrace A_1\right\rbrace = \Pr\left\lbrace A_1\right\rbrace\).
Inductive case: By induction, we have:
\(\Pr\left\lbrace ⋂_{i = 1}^n A_i\right\rbrace = ∏_{i = 1}^n \Pr\left\lbrace A_i \middle ⋂_{j = 1}^{i  1} A_j \right\rbrace\).
So
\(\Pr\left\lbrace ⋂_{i = 1}^{n + 1} A_i\right\rbrace\)
= \(\Pr\left\lbrace\left(⋂_{i = 1}^n A_i\right) \cap A_{n + 1}\right\rbrace\)
= \(\Pr\left\lbrace ⋂_{i = 1}^n A_i\right\rbrace ⋅ \Pr\left\lbrace A_{n + 1} \middle ⋂_{i = 1}^n A_i\right\rbrace\)
= \(\left(∏_{i = 1}^n \Pr\left\lbrace A_i \middle ⋂_{j = 1}^{i  1} A_j \right\rbrace\right) ⋅
\Pr\left\lbrace A_{n + 1} \middle ⋂_{i = 1}^n A_i\right\rbrace\)
= \(∏_{i = 1}^{n + 1} \Pr\left\lbrace A_i \middle ⋂_{j = 1}^{i  1} A_j \right\rbrace\).
C.26 ★
Describe a procedure that takes as input two integers a and b such that 0 < a < b and, using fair coin flips, produces as output heads with probability a / b and tails with probability (b  a) / b. Give a bound on the expected number of coin flips, which should be O(1). (Hint: Represent a / b in binary.)
Solution is implemented here.
The expected number of coin flips is 2.
C.27 ★
Show how to construct a set of n events that are pairwise independent but such that no subset of k > 2 of them is mutually independent.
Skipped.
C.28 ★
Two events A and B are conditionally independent, given C, if
Pr{A ∩ B  C} = Pr{A  C} ⋅ Pr{B  C}.
Give a simple but nontrivial example of two events that are not independent but are conditionally independent given a third event.
Skipped.
C.29 ★
You are a contestant in a game show in which a prize is hidden behind one of three curtains. You will win the prize if you select the correct curtain. After you have picked one curtain but before the curtain is lifted, the emcee lifts one of the other curtains, knowing that it will reveal an empty stage, and asks if you would like to switch from your current selection to the remaining curtain. How would your chances change if you switch? (This question is the celebrated Monty Hall problem, named after a gameshow host who often presented contestants with just this dilemma.)
The winning chance changes from 1 / 3 to 2 / 3.
C.210 ★
A prison warden has randomly picked one prisoner among three to go free. The other two will be executed. The guard knows which one will go free but is forbidden to give any prisoner information regarding his status. Let us call the prisoners X, Y, and Z. Prisoner X asks the guard privately which of Y or Z will be executed, arguing that since he already knows that at least one of them must die, the guard won’t be revealing any information about his own status. The guard tells X that Y is to be executed. Prisoner X feels happier now, since he figures that either he or prisoner Z will go free, which means that his probability of going free is now 1 / 2. Is he right, or are his chances still 1 / 3? Explain.
Still 1 / 3. Like the previous exercise, the chance of prisoner Z will go free is 2 / 3 after the guard says Y will be executed.
C.3 Discrete random variables
C.31
Suppose we roll two ordinary, 6sided dice. What is the expectation of the sum of the two values showing? What is the expectation of the maximum of the two values showing?
E[X + Y] = E[X] + E[Y] = 7 / 2 + 7 / 2 = 7.
According to equation C.25:
E[max(X, Y)]
= \(∑_{i = 1}^∞ \Pr\left\lbrace \max(X, Y) ≥ i\right\rbrace\)
= \(∑_{i = 1}^6 \Pr\left\lbrace \max(X, Y) ≥ i\right\rbrace\)
= \(∑_{i = 1}^6 \Pr\left\lbrace X ≥ i ∨ Y ≥ i\right\rbrace\)
= \(∑_{i = 1}^6 \Pr\left\lbrace X ≥ i\right\rbrace + \Pr\left\lbrace Y ≥ i\right\rbrace  \Pr\left\lbrace X ≥ i ∧ Y ≥ i\right\rbrace\)
= \(∑_{i = 1}^6 (7  i) / 6 + (7  i) / 6  (7  i)^2 / 36\)
= 161 / 36.
C.32
An array A[1‥n] contains n distinct numbers that are randomly ordered, with each permutation of the n numbers being equally likely. What is the expectation of the index of the maximum element in the array? What is the expectation of the index of the minimum element in the array?
E[arg max(A)] = (1 + n) / 2.
E[arg min(A)] = (1 + n) / 2.
C.33
A carnival game consists of three dice in a cage. A player can bet a dollar on any of the numbers 1 through 6. The cage is shaken, and the payoff is as follows. If the player’s number doesn’t appear on any of the dice, he loses his dollar. Otherwise, if his number appears on exactly k of the three dice, for k = 1, 2, 3, he keeps his dollar and wins k more dollars. What is his expected gain from playing the carnival game once?
E
=  Pr{k = 0} + \(∑_{i = 1}^3 i \Pr\left\lbrace k = i\right\rbrace\)
=  125 / 216 + \(∑_{i = 1}^3 i \binom{3}{i} (1 / 6) ^ i (5 / 6) ^ {3  i}\)
=  125 / 216 + 1 × 3 × 1 / 6 × 25 / 36 + 2 × 3 × 1 / 36 × 5 / 6 + 3 × 1 × 1 / 216 × 1
=  17 / 216.
So, do not gamble.
C.34
Argue that if X and Y are nonnegative random variables, then
E[max(X, Y)] ≤ E[X] + E[Y].
Since X and Y are nonnegative, E[max(X, Y)] ≤ E[X + Y] = E[X] + E[Y].
C.35 ★
Let X and Y be independent random variables. Prove that f(X) and g(Y) are independent for any choice of functions f and g.
Skipped.
C.36 ★
Let X be a nonnegative random variable, and suppose that E[X] is well defined. Prove Markov’s inequality:
Pr{X ≥ t} ≤ E[X] / t (C.30)
for all t > 0.
Skipped.
C.37 ★
Let S be a sample space, and let X and X′ be random variables such that X(s) ≥ X′(s) for all s ∈ S. Prove that for any real constant t,
Pr{X ≥ t} ≥ Pr{X′ ≥ t}.
Skipped.
C.38
Which is larger: the expectation of the square of a random variable, or the square of its expectation?
E[\(X^2\)]  \(\operatorname{E}^2[Y]\)
= E[\(X^2\)]  \(\operatorname{E}^2[X]\)
= Var[X]
≥ 0
C.39
Show that for any random variable X that takes on only the values 0 and 1, we have Var[X] = E[X] E[1  X].
Since X takes on only the values 0 and 1, we have \(X^2\) = X.
Var[X]
= E[\(X^2\)]  \(\operatorname{E}^2[X]\)
= E[X]  \(\operatorname{E}^2[X]\)
= E[X](1  E[X])
= E[X]E[1  X]
C.310
Prove that Var[a X] = \(a^2\) Var[X] from the definition (C.27) of variance.
Var[a X]
= E[\(a^2 X^2\)]  \(\operatorname{E}^2\left[a X\right]\)
= \(a^2\) E[\(X^2\)]  \(a^2 \operatorname{E}^2[X]\)
= \(a^2\)(E[\(X^2\)]  \(\operatorname{E}^2[X]\))
= \(a^2\) Var[X].