CSE 5311 Homework 1 Solution

Problem 2.2-1
Express the function n3 /1000 − 100n2 − 100n + 3 in terms of Θ-notation

Answer
Θ(n3 ).

Problem 2.3-3
Use mathematical induction to show that when n is an exact power of 2, the
solution of the recurrence
(
2 if n = 2,
T (n) =
2T (n/2) + n if n = 2k , for k > 1.

is T (n) = n lg n

Answer
First, establish a function for induction:

F (k) = T (2k ) (1)

We want to prove that:

F (k) = 2k lg 2k (2)

Base case:

F (1) = T (2) = 2 = 2 lg 2 = 21 lg 21 (3)

Prove for k + 1:

2k+1
F (k + 1) = T (2k+1 ) = 2T ( ) + 2k+1 (4)
2
= 2T (2k ) + 2k+1 = 2 · 2k lg 2k + 2k+1 (5)
k+1 k+1
=2 lg 2 (6)

1
Problem 2-1 Insertion sort on small arrays in
merge sort
Although merge sort runs in Θ(lg n) worst-case time and insertion sort runs
in Θ(n2 ) worst-case 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.

1. Show that insertion sort can sort the n/k sublists, each of length k, in
Θ(nk) worst-case time.
2. Show how to merge the sublists in Θ(n lg(n/k)) worst-case time.

3. Given that the modified algorithm runs in Θ(nk + n lg(n/k)) worst-case

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
Θ-notation?
4. How should we choose k in practice?

Answer
1. Sorting sublists
Note that sorting each list takes ak 2 + bk + c for some constants a, b and c. We
have n/k of those, thus:
n cn
(ak 2 + bk + c) = ank + bn + = Θ(nk)
k k

2. Merging sublists
Sorting a sublists of length k each takes:
(
0 if a = 1,
T (a) =
2T (a/2) + ak if a = 2p , if p > 0.
This makes sense, since merging one sublist is trivial and merging a sublists
means splitting dividing them in two groups of a/2 lists, merging each group
recursively and then combining the results in ak steps, since have two arrays,
each of length a2 k.
Proof by induction:
Base. Simple as ever:

T (1) = 1k lg 1 = k · 0 = 0
Step. Assume that T (a) = ak lg a and we calculate T (2a):

2
 T (2a) = 2T (a) + 2ak = 2(T (a) + ak) = 2(ak lg a + ak) = (7)
= 2ak(lg a + 1) = 2ak(lg a + lg 2) = 2ak lg(2a) (8)

This proves it. Now if we substitue the number of sublists n/k for a:
n n
T (n/k) = k lg = n lg(n/k)
k k
While this is exact only when n/k is a power of 2, it tells us that the overall
time complexity of the merge is Θ(n lg(n/k)).

3. The largest value of k

The largest value is k = lg n. If we substitute, we get:
n
Θ(n lg n + n lg ) = Θ(n lg n)
lg n
If k = f (n) > lg n, the complexity will be Θ(nf (n)), which is larger running
time than merge sort.

4. The value of k in practice

In practice, k should be the largest list length on which insertion sort is faster
than merge sort.

Problem 3-1 Asymptotic behavior of polynomials

Let
d
X
p(n) = a i ni
i=0

where ad > 0, be a degree-d polynomial in n, and let k be a constant. Use the

definitions of the asymptotic notations to prove the following properties.

1. If k ≥ d, then p(n) = O(nk ).

2. If k ≤ d, then p(n) = Ω(nk ).
3. If k = d, then p(n) = Θ(nk ).

4. If k > d, then p(n) = o(nk ).

5. If k < d, then p(n) = ω(nk ).

3
Answer
Let’s see that p(n) = O(nd ). We need do pick c = ad + b, such that:
d
X
= ad nd + ad−1 nd−1 + . . . + a1 n + a0 ≤ cnd
i=0

When we divide by nd , we get:

ad−1 ad−2 a0
c = ad + b ≥ ad + + 2 + ... + d
n n n
Or:
ad−1 ad−2 a0
b≥ + 2 + ... + d
n n n
If we choose b = 1, then we can choose n0 to be:
√ √
n0 = max(dad−1 , d ad−2 , . . . , d d a0 )
Now we have n0 and c, such that:

p(n) ≤ cnd for n ≥ n0

Which is the definition of O(n ). By chosing b = −1 we can prove the Ω(nd )
d

inequality and thus the Θ(nd ) inequality.

It’s very similar to prove the other inequalities.

Problem 3-2 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,  > 0, and c > 1 are constants. Your
answer should be in the form of the table with ”yes” or ”no” written in each
box.

Answer

A B O o Ω ω Θ
lgk n n yes yes no no no
nk cn yes yes no no no
√
n nsin n no no no no no
2n 2n/2 no no yes yes no
nlg c clg n yes no yes no yes
lg(n!) lg(nn ) yes no yes no yes

4
Problem 3-6 Iterated functions
We can apply the iteration operator ∗ used in the lg∗ function to any monoton-
ically increasing function f (n) over the reals. For a given constant c ∈ R, we
define the iterated function fc∗ by

fc∗ (n) = min{i ≥ 0 : f (i) (n) ≤ c}

which need not be well defined in all cases. In other words, the quantity fc∗ (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 fc∗ (n).

Answer
As shown in the table:

f (n) c fc∗ (n)

n−1 0 Θ(n)
lg n 1 Θ(lg∗ n)
n/2 1 Θ(lg n)
n/2 2 Θ(lg n)
√
n 2 Θ(lg lg n)
√
n 1 does not converge
1/3
n 2 Θ(log3 lg n)
n/ lg n 2 ω(lg lg n), o(lg n)

Problem 4.5-1
Use the master method to give tight asymptotic bounds for the following recur-
rences:

1. T (n) = 2T (n/4) + 1
√
2. T (n) = 2T (n/4) + n

3. T (n) = 2T (n/4) + n
4. T (n) = 2T (n/4) + n2

Answer
√
1. Θ(nlog4 2 ) = Θ( n)
√
2. Θ(nlog4 2 lg n) = Θ( n lg n)
3. Θ(n)

4. Θ(n2 )

5
Problem 4.5-4
Can the master method be applied to the recurrence T (n) = 4T (n/2) + n2 lg n?
Why or why not? Give an asymptotic upper bound for this recurrence.

Answer
With a = 4, b = 2, we have f (n) = n2 lg n 6= O(n2− ) 6= Ω(n2− ), so no - we
cannot apply the master method.
Let’s guess Θ(n2 lg2 n):

T (n) ≤ 4T (n/2) + n2 lg n (9)

2 2 2
≤ 4c(n/2) lg (n/2) + n lg n (10)
2 2 2
≤ cn lg(n/2) lg n − cn lg(n/2) lg 2 + n lg n (11)
2 2 2 2 2
≤ cn lg n − cn lg n lg 2 − cn lg(n/2) + n lg n (12)
2 2 2 2
≤ cn lg n + (1 − c)n lg n − cn lg(n/2) (c > 1) (13)
2 2 2
≤ cn lg n − cn lg(n/2) (14)
2
≤ cn2 lg n (15)

Exercise 4.6-2 is the general case for this.

Problem 4-1 Recurrence examples

Answer
Give asymptotic upper and lower bound for T (n) in each of the following re-
currences. Assume that T (n) is constant for n ≤ 2. Make your bounds as tight
as possible, and justify your answers.

1. T (n) = 2T (n/2) + n4
2. T (n) = T (7n/10) + n
3. T (n) = 16T (n/4) + n2
4. T (n) = 7T (n/3) + n2

5. T (n) = 7T (n/2) + n2
√
6. T (n) = 2T (n/4) + n
7. T (n) = T (n − 2) + n2

Problem 30-3
1. Θ(n4 ) (master method)
2. Θ(n) (master method, log10/7 1 = 0)

6
 3. Θ(n2 lg n) (master method)
4. Θ(n2 ) (master method)
5. Θ(nlog2 7 ) (master method)
√
6. Θ( n lgn ) (master method)
Pn/2
7. T (n) = n2 + T (n − 2) = n2 + (n − 2)2 + T (n − 4) = i=0 (n − 2i)
2
= Θ(n3 )

Problem 30-3
We can generalize the 1-dimensional discrete Fourier transform defined by
equa- tion (30.8) to d dimensions. The input is a d -dimensional array A =
(aj1 ,j2 ,··· ,jd ) whose dimensions are n1 , n2 , · · · , nd , where n1 n2 · · · nd = n. We
define the d-dimensional discrete Fourier transform by the equation
1 −1 n
nX X 2 −1 d −1
nX
yk1 ,k2 ,··· ,kd = ... aj1 ,j2 ,...,jd ωnj11k1 ωnj22k2 · · · ωnjddkd (16)
j1 =0 j2 =0 jd =0

for 0 ≤ k1 < n1 , 0 ≤ k2 < n2 , . . . , 0 ≤ kd < nd .

a. Show that we can compute a d-dimensional DFT by computing 1-dimensional

DFTs on each dimension in turn. That is, we first compute n/n1 sepa-
rate 1-dimensional DFTs along dimension 1. Then, using the result of
the DFTs along dimension 1 as the input, we compute n/n2 separate 1-
dimensional DFTs along dimension 2. Using this result as the input, we
compute n/n3 separate 1-dimensional DFTs along dimension 3, and so
on, through dimension d.
b. Show that the ordering of dimensions does not matter, so that we can
compute a d-dimensional DFT by computing the 1-dimensional DFTs in
any order of the d dimensions.
d. Show that if we compute each 1-dimensional DFT by computing the fast
Fourier transform, the total time to compute a d-dimensional DFT is
O(n lg n),independent of d.

Answer
a.
Reorder the summation as follows
1 −1 n
nX X 2 −1 d −1
nX
yk1 ,k2 ,··· ,kd = ... aj1 ,j2 ,...,jd ωnj11k1 ωnj22k2 · · · ωnjddkd (17)
j1 =0 j2 =0 jd =0
   
d −1
nX nd−1 −1 2 −1
nX 1 −1
nX
X
ωnjd kd ωnjd−1 kd−1 . . . ωnj2 k2 j1 k1


= d d−1 2
aj1 ,j2 ,...,jd ωn1  .
jd =0 jd−1 =0 j2 =0 j1 =0
(18)

Now each step of summation is single 1-dimensional DFT.

7
b.
With the result of a. and use the properties of summation and multiplication,
it can be proved that any two indexes s and t (1 ≤ s, t ≤ d) can be switch in
Equation 18. Hence, prove the hypothesis.

c.
The complexity of FFT on each dimension is

T (n, n1 ) = (n/n1 ) ∗ Θ(n1 lg n1 ) = Θ(n lg n1 ) (19)

T (n, n2 ) = (n/n2 ) ∗ Θ(n2 lg n2 ) = Θ(n lg n2 ) (20)
··· (21)
T (n, nd ) = (n/nd ) ∗ Θ(nd lg nd ) = Θ(n lg nd ) (22)

The overall complexity is

d
X
T (n) = T (n/ni ) (23)
i=1
d
X
= Θ(n lg ni ) (24)
i=1
= Θ(n lg n1 , n2 , · · · , nd ) (25)
= Θ(n lg n) (26)