Correction of Worksheet -1-

Exercise 1:
To answer this question, you must know some basic comparison scales.
for example, as 𝑛 → ∞ , and for constants a>1, b>0, and any real d, we have:
𝑑 < 𝑙𝑜𝑔𝑛 < 𝑛 < 𝑛𝑙𝑜𝑔𝑛 < 𝑛𝑎 < 𝑏 𝑛 < 𝑛! < 𝑛𝑛
You have also some allowed operations, for example,

• 𝑙𝑜𝑔𝑥 𝑦 = 𝑦𝑙𝑜𝑔𝑥
• 𝑙𝑜𝑔𝑥𝑦 = 𝑙𝑜𝑔𝑥 + 𝑙𝑜𝑔𝑦
1
• √𝑛 = 𝑛2
• Any polynomials dominate any logarithms: Generally, constant powers of n (𝑒𝑔. 𝑛𝑎 ) come after expressions
with only log factors. Even a small constant power of n grows faster than expressions with only logs. So √𝑛 =
1
𝑛2 grows faster than log 3 𝑛.
• if a>1 is a constant real and 1<n< n', then 𝑎n < 𝑎𝑛 ′ : Generally, expressions with some function of n in the
exponent go last. These are called "exponentials". Just see what has the biggest function in the exponent. For
example, since √𝑛 grows faster than logn, then 2√𝑛 grows faster than 2logn .

It is obvious that 𝑛 ! < (𝑛 +1)!

3𝑛+1 = 3 × 3𝑛 = 𝑂(3𝑛 ) (removing constant factors)
Since we have 𝑙𝑜𝑔𝑛 < 𝑛:
by adding log in both sides, we will have 𝑙𝑜𝑔𝑙𝑜𝑔(𝑛) < 𝑙𝑜𝑔𝑛.
1
1 1
And by multiplying both sides with 2 we will have 𝑙𝑜𝑔√𝑛 < √𝑛 (since 𝑙𝑜𝑔√𝑛 = log 𝑛2 = 2 𝑙𝑜𝑔𝑛 = 𝑂(𝑙𝑜𝑔𝑛 )
𝑙𝑜𝑔(𝑛 + 2𝑛3 ) ≈ log(2𝑛3 ) = 𝑙𝑜𝑔2 + 𝑙𝑜𝑔𝑛3 = 𝑙𝑜𝑔2 + 3𝑙𝑜𝑔𝑛 = 𝑂(𝑙𝑜𝑔𝑛) (Removing constant factors and lower order terms)
2

Functions by increasing asymptotic growth rate:

𝑙𝑜𝑔𝑙𝑜𝑔(𝑛) 𝑙𝑜𝑔√𝑛 √𝑛 𝑛2 + 𝑛(𝑙𝑜𝑔𝑛)3 4𝑛2 + 2𝑛3 3𝑛+1 𝑛! (𝑛 + 1)!
= 𝑂(𝑙𝑜𝑔𝑙𝑜𝑔(𝑛)) 𝑙𝑜𝑔(𝑛2 + 2𝑛3 ) 1 = 𝑂(𝑛2 ) = 𝑂(𝑛3 ) 3𝑛 = 𝑂(𝑛!) = 𝑂((𝑛 + 1)!)
= 𝑂(𝑛2 )
= 𝑂(𝑙𝑜𝑔𝑛) = 𝑂(3𝑛 )

Exercise 2:

• 𝑓1 (𝑛) = 𝑂(𝑛2 ) 𝑎𝑛𝑑 𝑓2 (𝑛) = 𝑂(𝑛) 𝑡ℎ𝑒𝑛 𝑓1 (𝑓2 (𝑛)) = 𝑂(𝑛3 ) ➔ True.
1) We have 𝑓1 (𝑛) = 𝑂(𝑛2 ), then based on the formal definition of the big O notation, there is a real constant 𝑐1 > 0 and an
integer constant 𝑛01 ≥ 1 such that 𝑓1 (𝑛) ≤ 𝑐1 𝑛2 , for 𝑛 ≥ 𝑛01 .
2) Similarly, we have 𝑓2 (𝑛) = 𝑂(𝑛), then there is a real constant 𝑐2 > 0 and an integer constant 𝑛02 ≥ 1 such that
𝑓2 (𝑛) ≤ 𝑐2 𝑛, for 𝑛 ≥ 𝑛02 .
𝑓1 (𝑓2 (𝑛)) ≤ 𝑓1 (𝑐2 𝑛), for 𝑛 ≥ 𝑛02 . (Adding 𝑓1 𝑜𝑛 𝑏𝑜𝑡ℎ 𝑠𝑖𝑑𝑒𝑠)
𝑛
Based on 1), we can write 𝑓1 (𝑓2 (𝑛)) ≤ 𝑓1 (𝑐2 𝑛) ≤ 𝑐1 (𝑐2 𝑛)2 , for 𝑛 ≥ 𝑛02 and c2 𝑛 ≥ 𝑛01  𝑛 ≥ 𝑐01
2
𝑛01
➔ 𝑓1 (𝑓2 (𝑛)) ≤ 𝑐1 𝑐22 𝑛2 , we can choose 𝑐 = 𝑐1 𝑐22 > 0 and 𝑛0 = max (𝑛02 , 𝑐2
).
Finally, based on the formal definition of the big O notation, 𝑓1 (𝑓2 (𝑛)) = 𝑂(𝑛2 ) and since 𝑂(𝑛2 ) < 𝑂(𝑛3 ) then
𝑓1 (𝑓2 (𝑛)) = 𝑂(𝑛3 ) , with 𝑂(𝑛3 ) is a non-tight bound.

• 𝑓1 (𝑛)𝑓2 (𝑛) = 𝑂(max (𝑓1 (𝑛), 𝑓2 (𝑛))) ➔ False.

Counter example: for 𝑓1 (𝑛) = 𝑛 𝑎𝑛𝑑 𝑓2 (𝑛) = n2
We have: 𝑓1 (𝑛)𝑓2 (𝑛) = 𝑛3 and 𝑂(max(𝑓1 (𝑛), 𝑓2 (𝑛))) = 𝑂(𝑛2 )
It is obvious that 𝑛3 is not 𝑂(𝑛2 ) since 𝑛3 dominates 𝑛2 .

Exercise3:
Exercice 4:
Initially, [5, 3, 8, 1, 9, 2, 11, 6]
For i = 1, [3, 5, 8, 1, 9, 2, 11, 6]
For i = 2, [3, 5, 8, 1, 9, 2, 11, 6]
For i = 3, [1,3, 5, 8, 9, 2, 11, 6]
For i = 4, [1, 3, 5, 8, 9, 2, 11, 6]
For i= 5, [1,2, 3, 5, 8, 9, 11, 6]
For i= 6, [1, 2, 3, 5, 8, 9, 11, 6]
For i = 7, [1, 2, 3, 5, 6, 8, 9, 11]
The worst-case time complexity is O(n2)

Exercise 5:
It is easier to find an element in a sorted array, so it is recommended to start with question 2).
Here I provided two possible solutions.
A better Algorithm with time complexity O(n), proposed by “Ons Kharrat Sophomore 2”:
Exercise 6:
In the worst case, we have the time complexity in the big O notation:
Linear search ➔ O(n)
Binary search ➔ O(logn)
Bubble sort ➔ O(n2)
1) Sorting the array with Bubble sort and then doing the searches using binary search:
We will sort the array once using bubble sort, then we will perform the binary search 100 times:
𝑓1 (𝑛) = 𝑂(𝑛2 ) + 100 × 𝑂(𝑙𝑜𝑔𝑛) ≈ (106 )2 + 102 × 20 = 1012 + 2 × 103

2) Just doing the searches with linear search (without sorting):

𝑓2 (𝑛) = 100 × 𝑂(𝑛) ≈ 102 × 106 = 108

➔𝑓2 (𝑛) ≪ 𝑓1 (𝑛)

Finally, it is clear that the second possibility (linear search without sorting) requires less work and time than the first one.