AMFAI Weekly Solution Set: Week 1

August 11, 2024

1. Here, f (n) = n log (n) + n log (log (n)) and g (n) = n log (n). We need to
show that ∃ c1 , c2 , n0 such that, ∀n ≥ n0 ,

c1 n log(n) ≤ n log(n) + n log(log(n)) ≤ c2 n log(n)

Consider the left-hand side,

c1 n log(n) ≤ n log(n) + n log(log(n))

Dividing by n on both sides,

c1 log(n) ≤ log(n) + log(log(n))

log(log(n))
∴ c1 ≤ 1 + (1)
log(n)
Similarly,
log(log(n))
c2 ≥ 1 + (2)
log(n)
∀n ≥ 2, ∃ c1 ≤ 1 and c2 ≥ 1 such that both equations, 1 and 2 are
satisfied. This is because the second term in both equations is a positive
value that slowly diminishes to 0. For example, consider, c1 = 1/2, c2 = 2
and n = 2. Then, we get 12 ≤ 1 ≤ 2.
2. Write, the first step,
 
2n n
T (n, n) = n + nT ,
3 2
Expand the recurrence term again,
 
2n n 4n n
T (n, n) = n + n + + +T ,
3 2 9 4
Thus, the general form of the recurrence looks like,
 k  k !
2 1
T n, n
3 2

1

k 2 k
It is easy to see that 21 n decreases much faster than


3 n meaning
that the base case y ≤ 50 will be reached first. Thus,
1 n
n ≤ 50 =⇒ k ≥ log
2k 50
Finally, we can write the general form as,

log( 50
n
)  k  k !
X 2 1
T (n, n) = n+ n + Θ(50)
3 2
k=0
log( 50
n
) k log( 50
n
) k
X 2 X 1
= n+ n + Θ(50) (1)
3 2
k=0 k=0

Plog( n )
k
Now, k=0 50 23 n is a GP series with a = n and r = 2/3. Thus, the
summation can be written as,

log( 50
n
)
 
k log( 50
n
)
X 2 1 − 32
n = n 2

k=0
3 1− 3

≈ 3n
k
Since, 21k < 32k , therefore we do not need to look at the second term in
equation 1. Finally, T (n, n) = Θ(3n) = Θ(n).
3. Every time that an array gets full, a new array is created and all elements
are copied. Since each time, the array size is doubled, we have, 2k =
n =⇒ k = log(n). Let the total element assignment cost be c = 0 When
|A0 | = 1, we can assign only x1 leading to c = 1 When x2 arrives, x1 needs
to be copied to A1 and x2 is added to the array, leading to c = 1 + 2 = 3.
Thus, the number of element assignments per array is simply equal to the
size of the array. Therefore, we have,
k
X
2j = 2j − 1 = Θ(n)
j=0

4. Given that n = ck =⇒ k = logc (n). Now, We first expand the given

recurrence.
n
f (n) = af + bnx (1)
n nc  n x
f = af 2 + b (2)
 nc   cn   cn x
f 2 = af 3 + b 2 (3)
c c c

2
 Replacing Eqn. 3 in 2 and Eqn. 2 in 1, we get,

a2
n  
3 x a
f (n) = a f 3 + bn 1 + x + 2x (4)
c c c

Based on the above pattern, we can write out the general form of the
recurrence as,
k−1
n X a j
f (n) = ak f + bnx
ck j=0
cx
logc (n)−1 
X a j
= alogc (n) d + bnx (5)
j=0
cx

If a = cx , then,

f (n) = alogc n d + bnx logc (n)

If a ̸= cx , then we first solve the summation term in Eqn. 5 which is a

GP series with the first term being 1, the common ratio being cax and the
number of elements being k.
k
X 1 − ax k 1 − (cak )x
k
k−1
!
1 − nax


c x
= = = −c (6)
j=0
1 − cax 1 − cax a − cx

Replacing the above equation in Eqn. 5, we get,

ak
!
1 − x
f (n) = ak d − bcx nx n
a − cx
bcx nx ak bcx
= ak d − +
a − cx a − cx
bcx bcx
   
logc n
= d+ a − nx
a − cx a − cx
bcx bcx
   
logc a
= d+ n − nx
a − cx a − cx

5. To prove that max(f (n), g(n)) = Θ(f (n) + g(n)) for asymptotically non-
negative functions f (n) and g(n), we need to show two things:
(a) max(f (n), g(n)) = O(f (n) + g(n))
(b) max(f (n), g(n)) = Ω(f (n) + g(n))
Once both are established, we can conclude that max(f (n), g(n)) = Θ(f (n)+
g(n)).
1. Prove max(f (n), g(n)) = O(f (n) + g(n))

3
By definition, max(f (n), g(n)) is the larger of the two functions. There-
fore, we have:
max(f (n), g(n)) ≤ f (n) + g(n)
To show max(f (n), g(n)) = O(f (n) + g(n)), we need to find constants
c > 0 and n0 such that for all n ≥ n0 :
max(f (n), g(n)) ≤ c(f (n) + g(n))

Since max(f (n), g(n)) ≤ f (n) + g(n), we can choose c = 1. Thus, we have:
max(f (n), g(n)) ≤ 1 · (f (n) + g(n))

Therefore, max(f (n), g(n)) = O(f (n) + g(n)).

2. Prove max(f (n), g(n)) = Ω(f (n) + g(n))

To show max(f (n), g(n)) = Ω(f (n) + g(n)), we need to find constants
c > 0 and n0 such that for all n ≥ n0 :
max(f (n), g(n)) ≥ c(f (n) + g(n))

Let’s use the following facts:

f (n) ≤ max(f (n), g(n))
g(n) ≤ max(f (n), g(n))

By adding these two inequalities, we get:

f (n) + g(n) ≤ max(f (n), g(n)) + max(f (n), g(n))
f (n) + g(n) ≤ 2 max(f (n), g(n))
Dividing both sides by 2:
1
(f (n) + g(n)) ≤ max(f (n), g(n))
2
This shows that max(f (n), g(n)) ≥ 21 (f (n)+g(n)), so we can choose c = 21 .
Therefore:
1
max(f (n), g(n)) ≥ (f (n) + g(n))
2
Thus, max(f (n), g(n)) = Ω(f (n) + g(n)).
Since we have shown both:
max(f (n), g(n)) = O(f (n) + g(n))
and
max(f (n), g(n)) = Ω(f (n) + g(n))
we conclude that:
max(f (n), g(n)) = Θ(f (n) + g(n))

4
6. We aim to prove that for any two functions f (n) and g(n), we have f (n) =
Θ(g(n)) if and only if f (n) = O(g(n)) and f (n) = Ω(g(n)).
To prove this, we will show both directions:
1. If f (n) = Θ(g(n)), then f (n) = O(g(n)) and f (n) = Ω(g(n)):

Assume f (n) = Θ(g(n)). By definition, this means there exist positive

constants c1 , c2 , and n0 such that for all n ≥ n0 :

c1 · g(n) ≤ f (n) ≤ c2 · g(n).

To show f (n) = O(g(n)): - From the inequality f (n) ≤ c2 · g(n), it

follows that f (n) does not grow faster than g(n) up to a constant factor.
- Therefore, f (n) = O(g(n)) with constant c = c2 .
To show f (n) = Ω(g(n)): - From the inequality f (n) ≥ c1 · g(n), it follows
that f (n) does not grow slower than g(n) up to a constant factor. -
Therefore, f (n) = Ω(g(n)) with constant c = c1 .
2. If f (n) = O(g(n)) and f (n) = Ω(g(n)), then f (n) = Θ(g(n)):
Assume f (n) = O(g(n)) and f (n) = Ω(g(n)). This means: - There exist
constants c1 and n0 such that for all n ≥ n0 :

f (n) ≥ c1 · g(n).

- There exist constants c2 and n1 such that for all n ≥ n1 :

f (n) ≤ c2 · g(n).

Let n′0 = max(n0 , n1 ). For n ≥ n′0 , we then have:

c1 · g(n) ≤ f (n) ≤ c2 · g(n).

Thus, f (n) is bounded both above and below by constant multiples of

g(n), which means f (n) = Θ(g(n)) with constants c1 and c2 .
In conclusion, f (n) = Θ(g(n)) if and only if f (n) = O(g(n)) and f (n) =
Ω(g(n)).
7. We aim to prove that if f = Ω(g) then f ∈
/ o(g).
Proof by Contradiction:

Assume, for the sake of contradiction, that f = Ω(g) and also f = o(g).
Definition of Ω(g)
By definition of Ω(g), there exist constants c > 0 and nΩ such that for all
n > nΩ ,
f (n) ≥ c · g(n). (7)

5
 Definition of o(g)
On the other hand, by definition of o(g), for any positive constant c, there
exists no such that for all n > no ,
f (n) < c · g(n). (8)

Consider n which is greater than both nΩ and no (e.g., n can be chosen

as max(nΩ , no ) + 1).
Then, according to the definition of Ω(g), we have:
f (n) ≥ c · g(n). (9)

And according to the definition of o(g), we have:

f (n) < c · g(n). (10)

However, the two inequalities (3) and (4) cannot both hold for the same
n. This is a contradiction.
Therefore, our initial assumption that f = Ω(g) and also f = o(g) must
be false. Hence, if f = Ω(g), then f ∈
/ o(g).
8. • We can use the limit definitions of o(n) and ω(n) to draw the same
conclusion.
f (n)
o(g(n)) = lim =0
n→∞ g(n)

and
f (n)
ω(g(n)) = lim =∞
n→∞ g(n)

Both of these cannot hold true as n approaches ∞.

Hence, no such f (n) exists, i.e., the intersection is indeed the empty
set.

Or

Function in o(g(n)):
If f (n) ∈ o(g(n)), then for every positive constant c1 > 0 and for
sufficiently large n, we have:
0 ≤ f (n) < c1 · g(n)

• Function in ω(g(n)):
If f (n) ∈ ω(g(n)), then for every positive constant c2 > 0 and for
sufficiently large n, we have:
0 ≤ c2 · g(n) < f (n)

6
 • Combine the Results:
Suppose there is a function f (n) that belongs to both o(g(n)) and
ω(g(n)). Then for some constants c1 > 0 and c2 > 0, we need:

c2 · g(n) < f (n) < c1 · g(n)

This is a contradiction because f (n) cannot be simultaneously greater

than c2 · g(n) and less than c1 · g(n) for all sufficiently large n. As n
grows, c2 · g(n) and c1 · g(n) can be made arbitrarily close by choosing
appropriate c1 and c2 , making it impossible for f (n) to satisfy both
conditions.
Therefore, no such function f (n) can exist, which implies:

o(g(n)) ∩ ω(g(n)) = ∅

This completes the proof that the intersection of o(g(n)) and ω(g(n)) is
indeed the empty set.

9. (a) T (n/2) + c
10. The value of j gets updated as j = 2, 4, 16, · · · . The general form of this
p p
series is 22 . Thus, the number of terms in this series are, 22 = n =⇒
p = log(log(n)). The outer loop runs for n times. Thus, the total time
complexity can be written as Θ(n log(log(n))).