1

Efficient Algorithms and Intractable Problems

Student’s Name
Institution
Course
Professor
Due Date
 2

(a) Simplify the following expressions into a single logarithm (i.e. in the form logab):

ln x
(i) ln y is simplified as logyx

(ii) ln x + ln y
ln(xy)=ln(x)+ln(y)
ln(x)= logex
ln(y)=logey
ln(xy)=logex+logey=loge(xy)

(iii) ln x − ln y
ln x
ln x − ln y = ln y

This is simplified to logyx

(iv) 170 ln x
170ln(x)=ln(x170)=loge(x170)

b) Give a simple proof for each of the following identities:

1
a) Xlog1/x y= 𝑦
Let a= log1/xy
1
Then (𝑥)a=b
Substitute into the first equation:
Xlog1/xb= xa
1 1
Since (𝑥)a=b, then xa = 𝑏
Therefore the simplified expression is:
1
Xlog1/xb= 𝑏
Thus:
1
Xlog1/x y= 𝑦

𝑛(𝑛+1)
b) ∑𝑛𝑖=1 𝑖 = 2
 3

Let the identity holds for some arbitrary positive integer a. That is,
𝑎(𝑎+1)
∑𝑎𝑖=1 𝑖 =
2
Show that it also holds for a+1:
(𝑎+1)(𝑎+2)
∑𝑎+1
𝑖=1 𝑖 = 2
Simplifying starting from the left hand side:
𝑎(𝑎+1)
∑𝑎+1 𝑎
𝑖=1 𝑖 = ∑𝑖=1 𝑖 + (a+1) = + (a+1)
2

Combine the terms with a common denominator:

𝑎(𝑎+1)+(𝑎+1) (𝑎+1)+2(𝑎+1) (𝑎+1)(𝑎+2)
= =
2 2 2

Therefore, the identity holds for all positive integers n.

1−𝑟 𝑛+1
𝑎 ( 1−𝑟 ) , 𝑟 ≠ 1
c) ∑𝑛𝑘=0 ar 𝑘
={
𝑎(𝑛 + 1), 𝑟 = 1
To prove the given identity for the geometric series, let's consider two cases: 𝑟 ≠
1 and r=1.
The sum of a geometric series is given by the formula:
1−𝑟 𝑛+1
Sn= 𝑎 ( )
1−𝑟
Verifying the identity:
∑𝑛𝑘=0 ar K = a + ar + ar2 + …+arn
Using the formula for the sum of a geometric series:
1−𝑟 𝑛+1
Sn= 𝑎 ( )
1−𝑟

Let the formula hold for n=k:

1−𝑟 𝑘+1
Sk= 𝑎 ( )
1−𝑟

Let’s prove it for n = k+1

1−𝑟 (𝑘+1)+1
Sk+1= 𝑎 ( )
1−𝑟

1−𝑟 𝑘+2
Sk+1= 𝑎 ( )
1−𝑟

1−𝑟 𝑘+2
Sk+1= 𝑎 ( )
1−𝑟

Expressing Sk+1 in terms of Sk

1−𝑟 𝑘+2
Sk+1= 𝑎 ( )
1−𝑟

Sk+1 = Sk . r
 4

Substituting the formula for Sk:

1−𝑟 𝑘+1
Sk+1= 𝑎. ( ).r
1−𝑟

𝑟−𝑟 𝑘+2
Sk+1= 𝑎 ( )
1−𝑟

𝑟−𝑟 𝑘+2
Sk+1= 𝑎 ( )
1−𝑟

Thus, the formula holds for n=k+1.

This proves the identity holds when 𝑟 ≠ 1

Now, substitute Sn back into the expression:

1−𝑟 𝑛+1
∑𝑛𝑘=0 ar 𝑘 = 𝑎 ( )
1−𝑟
This proves the identity for the case r ≠ 1
For case 2 when r=1:
When r=I , the sum of the geometric series becomes :
∑𝑛𝑘=0 ar 𝑘 =∑𝑛𝑘=0 a
The sum of n+1 terms of the constant a is a(n+1), therefore:
∑𝑛𝑘=0 𝑎𝑟 𝑘 = a +a +a +……a(n+1 terms)
Simplifies to:
∑nk=0 ar k = a(n+1)
This proves the identity for the case r=1.
So, the identity holds for both cases, and we've successfully proved the formula
for the sum of a geometric series.

5 Recurrence Relations
For each part, find the asymptotic order of growth of T; that is, find a function g such that
T(n) = Θ(g(n)). Show your reasoning and do not directly apply the Master Theorem;
doing so will yield 0 credit.
In all subparts, you may ignore any issues arising from whether a number is an integer.
(a) T(n) = 2T(n/3) + 5n
To find the asymptotic order of growth of the given recurrence relation T(n)=2T(n/3)+5n, the
Master Theorem is used.
The recurrence relation is in the form:
 5

𝑛
T(n) = aT (𝑏 )+f(n)

where:
 a is the number of sub problems,
 b is the factor by which the problem size is reduced,
 f(n) is the cost of the work done outside the recursive calls.
The Master Theorem has the following form:
𝛩(𝑛𝑙𝑜𝑔𝑏𝑎 ) 𝑖𝑓 𝑓(𝑛) = 𝑂(𝑛𝑙𝑜𝑔𝑏𝑎−£ ) for some 𝜖 > 0
T(n) = { 𝛩(𝑓(𝑛)) 𝑖𝑓 𝑓(𝑛) = 𝛩(𝑛𝑙𝑜𝑔𝑏𝑎 )
𝛩(𝑓(𝑛)𝑙𝑜𝑔𝑏𝑎 ) 𝑖𝑓 𝑓(𝑛) = Ω(𝑛𝑙𝑜𝑔𝑏𝑎+£ ) 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝜖 > 0
For T(n) = 2T(n/3) + 5n
 a=2 (two sub problems),
 b=3 (problem size reduced by a factor of 3),
 f (n)=5n.

Comparing f(n) with nlogba ,

f(n)= 5n = 𝑂(𝑛𝑙𝑜𝑔3 2−£ )
Since f(n) is 𝑂(𝑛𝑙𝑜𝑔𝑏𝑎−£ ) for some 𝜖 > 0 , that is in case 1 of the master theorem.
Therefore, the solution for the recurrence relation T(n) is:
T(n) = 𝛩(𝑛𝑙𝑜𝑔𝑏𝑎 ) = Θ (𝑛𝑙𝑜𝑔3 2 )
So, the asymptotic order of growth for T(n) is 𝑛𝑙𝑜𝑔3 2

(b) T (n) = 169T (n/170) + Θ (n)

The recurrence relation is in the form:
𝑛
T(n) = aT (𝑏 )+f(n)

Where a= 169, b = 170, and f(n) = Θ (n)

Substituting in 𝑙𝑜𝑔𝑏 𝑎:
𝑙𝑜𝑔170 169 ≈ 0.9995
Compare f (n) with 𝑛𝑙𝑜𝑔𝑏𝑎 :
 6

f (n)=Θ(n)=Θ(𝑛𝑙𝑜𝑔170169 )
Since 𝑓(𝑛) = 𝛩(𝑛𝑙𝑜𝑔𝑏𝑎 ) as in case 2 of the master theorem. Therefore, the solution for the
recurrence relation T(n) is :
T(n)=Θ(f(n)logn)=Θ(nlogn)
the asymptotic order of growth for T(n) is Θ(n log n).

(c) An algorithm A takes Θ (n2) time to partition the input into 5 sub-problems of size n/5
each and then recursively runs itself on 3 of those sub problems. Describe the recurrence
Relation for the run-time T (n) of A and find its asymptotic order of growth.
The algorithm A can be described with the following recurrence relation:
𝑛
T(n) = Θ(n2)+ 3T(5)

Here, the Θ(n2) term represents the time taken to partition the input into 5 sub-problems of size
n/5 each.
𝑛
The recursive part 3T (5) represents the algorithm running itself on three of those sub-problems.
𝑛
Each sub-problem has size ( 5), and there are three of them, so we multiply by 3.

Analyze the asymptotic order of growth using the Master Theorem. The recurrence relation is in
the form:
𝑛
T(n) =Θ ( f(n)) + aT (𝑏 )

Where f (n) =n2, a=3, and b=5.

Substituting into 𝑙𝑜𝑔𝑏 𝑎 :
𝑙𝑜𝑔5 3
Compare f(n) with 𝑛𝑙𝑜𝑔𝑏𝑎 :
n2= O ( 𝑛𝑙𝑜𝑔5 3−£ ):
Since f(n) is O ( 𝑛𝑙𝑜𝑔𝑏𝑎−£ ) for some 𝜖 > 0 , as in the case 1 of the master theorem.
Therefore the solution for the recurrence relation T(n) is :
T(n) = Θ( 𝑛𝑙𝑜𝑔𝑏𝑎 ) = Θ ( 𝑛𝑙𝑜𝑔5 3 ):
T(n) = Θ( f(n)log n) = Θ(n2 log n)
The asymptotic order of growth for the run time of algorithm A is Θ (n2log n ).
 7

(d) T(n) = 6T(n/6) + Θ(n)

The recurrence relation is in the form:
𝑛
T(n) = aT (𝑏 )+f(n)

Where a=6, b=6, and f(n)=Θ(n)

Comparing f(n) with 𝑛𝑙𝑜𝑔𝑏𝑎 :
f (n)=Θ(n)=Θ(𝑛𝑙𝑜𝑔6 6 )
Since 𝑓(𝑛) = 𝛩(𝑛𝑙𝑜𝑔𝑏𝑎 ) as in case 2 of the master theorem. Therefore, the solution for the
recurrence relation T(n) is :
T(n)=Θ(f(n)logn)=Θ(nlogn)
The asymptotic order of growth for T (n) is Θ(n log n).
(e) T(n) = T(3n/5) + T(4n/5) (We have T(1) = 1)
Hint: first, compute a reasonable upper and lower bound for T(n). Then, try to guess
a T(n) of the form anb and then use induction to argue that it is correct.
Using upper and lower bounds for T(n):
Upper bound:
A solution of T(n) = O(nb), where b is a constant.
Substituting this into the recurrence relation we get:
3𝑛 4𝑛 3𝑛 4𝑛
T(n)≤ T( 5 ) + T( 5 ) ≤ c(( 5 )b+( 5 )b)

Simplifying:
3𝑏 4𝑏
T(n)≤c(5𝑏 + 5𝑏 )nb

Let's choose b such that 3b+4b=5b. A reasonable choice is b=1. Therefore, we assume and
conclude T(n)=O(n).
Lower bound:
Let T(n) = Ω(na) for some constant a.
Substituting this into the recurrence relation:
3𝑛 4𝑛 3𝑛 4𝑛
T(n)≥ T( 5 ) + T( 5 ) ≥ c(( 5 )a+( 5 )a)
3𝑎 4𝑎
T(n)≥c(5𝑎 + 5𝑎 )na
 8

Let's choose a such that 3a+4a=5a. A reasonable choice is a=2. Therefore, we assume and
conclude T(n)= Ω(n2).
Combining the upper and lower bounds we get:
Since both have T(n)=O(n) and T(n)= Ω(n2) we conclude that T(n)= Θ (n).
Induction process:
Let T(k) = ck for all k <n .
Show T(n) = cn holds.
3𝑛 4𝑛 3𝑛 4𝑛
T(n)= T( 5 ) + T( 5 ) ≤ c( 5 )+( 5 )=cn

The base case T (1) =1 is satisfied by choosing c=1

Therefore, by induction, T(n) = Θ(n)

6 In Between Functions
In this problem, we will find a function f(n) that is asymptotically worse than polynomial
time but still better than exponential time. In other words, f has to satisfy two properties,
• For all constants k > 0, f(n) = Ω(nk) (1)
• For all constants c > 0, f(n) = O(2cn) (2)
(a) Try setting f(n) to a polynomial of degree d, where d is a very large constant. So
f(n) = a0 + a1n + a2n2· · · + adnd . For which values of k (if any) does f fail to satisfy(1)?
Let f(n)=a0+a1n+a2n2+…+adnd, where d is a very large constant.
Finding for which values of k does f(n)=Ω(nk) fail.
For any positive k, as n becomes large, the term adnd dominates all other terms in the polynomial.
Therefore,f(n) is Ω(nd), but it fails for k<d.

(b) Now try setting f(n) to an, for some constant a > 1 that’s as small as possible while
still satisfying (1) (e.g. 1.000001). For which values of c (if any) does f fail to satisfy
(2)?
Hint: Try rewriting an as 2bn first, where b is a constant dependent on a.
So far we have found that the functions which look like O(nd) for constant d are too
 9

small and the functions that look like O(an) are too large even if a is a tiny constant.

Let f(n)=an for some constant a>1.

Finding for which values of c does f(n)=O(2cn) fail.

Rewriting an as 2(𝑛𝑙𝑜𝑔2 𝑎) .
1 1
f (n)=O(2cn) holds for any c>(𝑙𝑜𝑔 . However It fails when c ≤ (𝑙𝑜𝑔
2 𝑎) 2 𝑎)

(c) Find a function D(n) such that setting f(n) = O(nD(n)) satisfies both (1) and (2). Give a proof
that your answer satisfies both.
Hint: make sure D(n) is asymptotically smaller than n.
Let D(n) = 𝑙𝑜𝑔2 𝑛 .

Let f(n) = 2(𝑙𝑜𝑔2 𝑛) = n

Proof:
1. Satisfying (1):
For all constants k>0, f(n)=n=Ω(nk) because nk grows slower than n for any k>0
2. Satisfying (2):
For all constants c>0, f(n) = 2(𝑙𝑜𝑔2 𝑛) = n is O(2cn) since the growth rate of n is linear.
Thus f(n)= O(nD(n)) with D(n) = 𝑙𝑜𝑔2 𝑛 satisfies both conditions.
10