Scribd
Upload
56 views

CSE 101 Homework 0 Solutions: Winter 2021

This document contains solutions to four questions about algorithm analysis. Question 1 analyzes the runtime of two algorithms and determines they are Θ(n^2) and Θ(n log n), respectively. Question 2 sorts five functions by asymptotic growth rate and determines which are polynomial. Question 3 proves that if there is a walk between two vertices in a graph, there is also a walk where all vertices are distinct. Question 4 analyzes a recurrence relation, determining an exact solution, Θ expression, and identifying a flaw in an attempted inductive proof.

Uploaded by

Hoang Nguyen
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
56 views

CSE 101 Homework 0 Solutions: Winter 2021

This document contains solutions to four questions about algorithm analysis. Question 1 analyzes the runtime of two algorithms and determines they are Θ(n^2) and Θ(n log n), respectively. Question 2 sorts five functions by asymptotic growth rate and determines which are polynomial. Question 3 proves that if there is a walk between two vertices in a graph, there is also a walk where all vertices are distinct. Question 4 analyzes a recurrence relation, determining an exact solution, Θ expression, and identifying a flaw in an attempted inductive proof.

Uploaded by

Hoang Nguyen
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

CSE 101 Homework 0 Solutions

Winter 2021

Question 1 (Program Runtimes, 20 points). Consider the following programs:


Alg1(n):
For i = 1 to n
j = 1
while j < n
j = j+1

Alg2(n):
For i = 1 to n
j = 1
while j < n
j = j+i

For each of these algorithms, compute the asymptotic runtime in the form Θ(−).
Solution 1.
Alg1: Each iteration of the for loop involves Θ(n) operations, as j is always incremented n − 1 times. There
are n iterations of the for loop, so our total runtime is n ∗ Θ(n) = Θ(n2 ) .
n
Alg2: In each iteration of the for loop, Pnj is incremented
Pn Θ( i )1 times. i iterates from 1 to n, th
so our total
n n n
runtime is Θ(n) + Θ( 2 ) + Θ( 3 ) + · · · = i=1 Θ( i ) = n · i=1 Θ( i ). The summation is just the n harmonic
number Hn , and Hn = Θ(log n), so our total runtime is Θ(n log n) .

Question 2 (Big-O Computations, 20 points). Sort the functions below in terms of their asymptotic growth
rates. Which of these functions have polynomial growth rates? Remember to justify your answers.
• a(n) = n + n1/2 + n1/3 + . . . + n1/n
• b(n) = 3dlog2 (n)e
• c(n) = n2 (2 + cos(n))
• d(n) = n100 2n/2
• e(n) = 2n
Solution 2.

• n ≤ a(n) ≤ n + log(n)n1/2 + n × n1/ log(n) = O(n) since log(n) ∈ O(n1/2 ). Notice that l = n1/ log(n) is
a constant since log(l) = 1/ log(n) ∗ log(n) = 1 =⇒ l = e. So a(n) ∈ θ(n).
• b(n) = 3dlog2 (n)e ≤ 3 × 3log2 (n) = 3 × 2log2 (3) log2 (n) = 3nlog2 (3) and b(n) = 3dlog2 (n)e ≥ 3log2 (n) /3 =
2log2 (3) log2 (n) /3 = nlog2 (3) /3. So b(n) = θ(nlog2 (3) ).
• Ω(n2 ) ≤ c(n) = n2 (2 + cos(n)) ≤ O(n2 ) since 1 ≤ 2 + cos(n) ≤ 3.
• Ω(n2 ) ≤ d(n) = n100 2n/2

1
• Ω(n100 2n/2 ) ≤ e(n) = 2n since n100 ∈ O(2n/2 )

So the order is clearly a, b, c, d, e by reasons above. The functions a, b, c are polynomial growth rates. The
functions d, e are not polynomial growth since they are greater than 2n/2 which is not polynomial growth.

Question 3 (Walks and Paths, 30 points). In a graph G we say that there is a walk from vertex u to another
vertex w if there is a sequence of vertices u = v0 , v1 , . . . , vn = w so that (vi , vi+1 ) is an edge of G for each
0 ≤ i < n. Prove that if there is a walk from u to w there is a walk where all of the vertices vi are distinct.
Hint: if two are the same show how you can use this to construct a shorter walk.

Solution 3.
Suppose we have a walk from vertex u to vertex w, the walk is denoted as (u, v0 , v1 , ..., vn , w) = p where
there are duplicated vertices in p. Assume that we cannot find a walk p̂ such that p̂ ⊂ p and ele-
ments in p̂ are distinct. Suppose vi , vj ∈ p, vi = vj , i < j. Then (vi , vj+1 ) is also an edge of G. So
p̂ = (u, v0 , ..., vi , vj+1 , ..., vn , w) is a valid walk. We can apply this procedure until all the elements are
distinct. This contradicts with the assumption we have. So there exists a walk from v to w where all the
vertices vi are distinct.
Question 4 (Recurrence Relations, 30 points). Consider the recurrence relation

T (1) = 1, T (n) = 2T (bn/2c) + n.

(a) What is the exact value of T (2n )?


(b) Give a Θ expression for T (n). Hint: compare its value to that at nearby powers of 2.
(c) Consider the following purported proof that T (n) = O(n) by induction:
If n = 1, then T (1) = 1 = O(1).
If T (m) = O(m) for m < n, then

T (n) = 2T (bn/2c) + n = O(n) + O(n) = O(n).

Thus, T (n) = O(n).


What is wrong with this proof ? Hint: consider the implied constants in the big-Os.

Solution 4.

(a) We show by induction that ∀n ≥ 0, T (2n ) = (n + 1) · 2n . For n = 1, clearly T (20 ) = 1 = (0 + 1) · 1.


Suppose the relationship holds for all n ≤ t. Then, T (2n+1 ) = 2T (2n ) + 2n+1 = 2(n + 1) · 2n + 2n+1 =
(n + 2) · 2n+1 .

(b) Let k = blog2 nc Observe that T (2k ) ≤ T (n) ≤ T (2k+1 ) or

(blog2 nc + 1)2blog2 nc ≤ T (n) ≤ (blog2 nc + 2)2blog2 nc+1

implying
log2 n · n/2 ≤ T (n) ≤ 2(log2 n + 2)n.

Thus T (n) = Θ(n log2 n) .

(c) In order for the induction step to be correct, we require that the same constant as that of T (m) for
m < n is also obtained for T (n). Under the assumption that ∀m < n, T (m) = O(m), let c be a constant
such that for a sufficiently large m, T (m) ≤ c·m. This only implies that T (n) = 2T (bn/2c)+n ≤ (c+1)n
and the induction step fails.

You might also like