Recurrence Relation

Recurrence Relation
A recurrence relation is an equation that recursively defines a
sequence where the next term is a function of the previous terms.
For example, Fibonacci series
Fn = Fn-1+Fn-2
 Recurrence Relation
void fun(int n) fun (3)
{
if (n > 0) {
cout << n << " "; 3 fun(2)
fun(n - 1);
}
} 2 fun(1)
int main()
{
fun(3); 1 fun(0)
return 0;
} stop

Recursion Tree
Time=?
It can be observed from the recursion tree that there are 3 self call, and, in each call, it
takes 1 unit of time for printing. So, time taken for n=3 will 3 and for n=n, it will be n,
i.e., O(n).
Recurrence Relation
void fun(int n) T(n)
{
if (n > 0) {
cout << n << " "; 1
fun(n - 1); T(n-1)
}
}
int main()
{
fun(3);
return 0;
}

Recurrence relation

T(n)= 1+ T(n-1) if n>0

1 n=0
Solving Recurrence Relation
Recurrence relation T(n-1) = 1+T(n-2)----(i)
1+ T(n-1) if n>0 T(n-2) = 1+T(n-3)----(ii)
T(n)= 1 n=0

Use Back substitution method T(n) = k + T(n-k) --------------- (iii)

Substitute T(n-1) in main equation As we are subtracting 1 from n every

time, so at one point it will reach 0.
T(n)= [1+T(n-2)]+1 Assume n-k=0
= 2+T(n-2) So, n=k
Replace k by n in equation (iii)
If we substitute T(n-2) in main equation, then
T(n) = n + T(n-n)
T(n)= [1+T(n-3)]+2 T(n) = n+ T (0)
= 3+T(n-3) T(n) = n+1
.
. T(n) = O(n)
T(n) = k + T(n-k)
Recurrence Relation
void fun(int n) T(n)
{
if (n > 0) { 1
for(i=0; i<n; i++) n+1
cout << n << " "; n
fun(n - 1); T(n-1)
}
}

Recurrence relation

n+ T(n-1) if n>0
T(n)= 1 n=0
 Solving Recurrence Relation
Recurrence relation

n+ T(n-1) if n>0
T(n)= 1 n=0
Recursion tree method

T(n) n
n T(n-1) n-1
n-1 T(n-2) n-3
n-2 T(n-3)
.
T(2) 2
Time taken 2 T(1) 1
1 T(0)
0+1+2+3+…+n-1+n
= n(n+1)/2 stop
= O(n2)
 Solving Recurrence Relation
Recurrence relation
Use back substitution method
n+ T(n-1) if n>0 T(n-1) = T(n-2) + n-1 -(ii)
T(n)= 1 n=0 T(n-2) = T(n-3)+ n-2 -(iii)

T(n) = T(n-1) + n -(i)

Replace value of T(n-1) in eq. (i)

T(n)= [T(n-2)+n-1]+n
= T(n-2)+n-1+n
Replace value of T(n-2) in eq. (i)

T(n) = T(n-3)+n-2+n-1+n
.
T(n) = T(n-k)+(n-(k-1))+ (n-(k-2))+ … + (n-1) + n

At one time n-k will reach zero

n-k=0 => n=k
T(n)= T(n-n) + (n-(n-1)) + (n-(n-2)) + …+n-1 +n
= T(0)+1+2+3+…+n-1+n
= n(n+1)/2
= O(n2)
Recurrence Relation
void fun(int n) T(n)
{
if (n > 0) {
for(i=0; i<n; i=i*2) log n
cout << n << " "; log n
fun(n - 1); T(n-1)
}
}

Recurrence relation

log n+ T(n-1) if n>0

T(n)= 1 n=0
 Recurrence Relation
Recurrence relation
T(n)
log n+ T(n-1) if n>0
T(n)= 1 n=0
log n T(n-1)

Time taken
log n-1 T(n-2)
0+log1+log2+log3+…+logn-1+logn
= log[1x2x3x…xn-1xn]
= log (n!) log n-2 T(n-3)
The upper bound for n! is nn .
= log (nn) T(2)
= nlog n
log 2 T(1)

Use back substitution method log 1 T(0)

and find the time complexity
Stop
 Recurrence Relation
Observation about recurrence relation for decreasing function

T(n) = T (n-1) +1 O(n)

T(n) = T(n-1) + n O(n2)
T(n) = T(n-1) + log n O(n log n)

It can be observed from the above that first recurrence is repeated n times
so, time will be O(n). Similarly, for the third one log n is repeated n times
so, time taken will be n log n.
T(n) = T (n-1) + n2 ?
T(n) = T (n-2) + 1 ?
T(n) = T (n-100) + n ?

What will be time if some constant is being multiplied to decreasing function?

T(n) = 2T (n-1) + 1 ?
Recurrence Relation
Recurrence relation

2T(n-1)+1 if n>0
T(n)= 1 n=0

Use Back substitution method

T(n)= 2T(n-1)+1 (i)

T(n)= 2[2T(n-2)+1]+1
= 22 T(n-2)+2+1 (ii)

T(n) = 22 [2T(n-3)+1]+2+1
= 23 T(n-3)+ 22+2+1 (iii)

T(n) = 2k T(n-k)+ 2k-1+…+2+1 (iv)

Assume n-k=0 => n=k

T(n) = 2n T(0)+ 2n-1+ 2n-2…+2+1
= 2n + 2n-1+ … +22+2+1
= 2n+1
T(n)= O(2n)
Recurrence Relation
Observation about recurrence relation for decreasing function
T(n) = T (n-1) +1 O(n)
T(n) = T(n-1) + n O(n2)
T(n) = T(n-1) + log n O(n log n)
T(n) = 2T(n-1) +1 O(2n )
T(n) = 3T(n-1) +1 O(3n )
T(n) = 2T(n-1) +n O(n2n )
Master theorem for decreasing function
From the above analysis, we can write-
T(n) = aT(n-b) +f(n), where a>0, b>0, and f(n)= O(nk ), where k>=0
If a=1, then
T(n) = O(nk+1) or O(n*f(n))
If a>1, then
T(n) = O(an/b nk) or O(an/b *f(n))
If a<1, then
T(n) = O(nk) or O(f(n))
Recurrence Relation
Recurrence relation for dividing functions
T(n/2)+1 if n>0
T(n)= 1 n=1

Use Back substitution method

T(n)= T(n/2)+1 (i)
T(n)= [T(n/2*2)+1]+1 T(n)= O(log n)
= T(n/22 )+2 (ii)
T(n) = [T(n/ 23)+1]+2
= T(n/ 23)+3 (iii)
T(n) = T(n/ 2k)+k (iv)
Assume n/ 2k =1 => 2k =n T(n/2)+n if n>0
T(n) = T(1) +k (v) T(n)= 1 n=1
Take log
k log22 =log n
K = log n (vi) T(n)=?
Put value of k in equation (v)
Recurrence Relation
Recurrence relation for dividing functions
T(n/2)+n if n>0
T(n)= 1 n=1
Use Back substitution method
T(n)= T(n/2)+n (i)
T(n)= [T(n/2*2)+n/2]+n
= T(n/22 )+n/2+n (ii)
T(n) = [T(n/ 23)+n/4]+n/2+n
= T(n/ 23)+n/4+n/2+n (iii)
T(n) = T(n/ 2k)+ n/ 2k-1 +n/ 2k -2+…+n/2+n (iv)
Assume n/ 2k =1 => K = log n
T(n) = T(1)+ n[1/ 2k-1 +1/ 2k -2+…+1/2+1] (v)
T(n) = T(1)+ n[1+1]
T(n) = 1+2n
T(n)= O(n)
Recurrence Relation
Recurrence relation for dividing functions
2T(n/2)+n if n>0
T(n)= 1 n=1
Use Back substitution method
T(n)= 2T(n/2)+n (i)
T(n)= 2[2T(n/2*2)+n/2]+n
= 22 T(n/22 )+n + n (ii)
T(n) = 22[2T(n/ 23)+n/4]+2n
= 23T(n/ 23)+n+2n (iii)
T(n) = 2kT(n/ 2k)+ kn (iv)

Assume n/ 2k =1 => n= 2k and k = log n

T(n) = n*T(1)+ nlogn (v)
T(n) = n+ nlogn

T(n)= O(nlog n)
Recurrence Relation
Masters Theorem for Dividing Functions
T(n) = aT(n/b) +f(n), where a>=1, b>1, and f(n)= O(nk logpn), where k>=0

Determine values of logba and k. Based on the values of k and logba, there
will be three cases.

Case 1: if logba > k, then T(n)=O(nlogba)

Case 2: if logba = k, and

(i) if p>-1, then T(n)=O(nk logp+1n)
(ii) if p=-1, then T(n)=O(nk loglogn)
(iii) if p<-1, then T(n)=O(nk)

Case 3: if logba < k, and

(i) if p>=0, then T(n)=O(nk logpn)
(ii) if p<0, then T(n)=O(nk)
 Recurrence Relation
Use Masters Theorem and find time complexity
1. T(n)= 2T(n/2)+1 T(n) = aT(n/b) +f(n), where a>=1, b>1, and
f(n)= O(nk logpn), where k>=0
2. T(n)= 4T(n/2)+n
Case 1: if logba > k, then O(nlogba)
3. T(n)= 8T(n/2)+n2
Case 2: if logba = k, and
4. T(n)= 8T(n/2)+ n logn
(i) if p>-1, then O(nk logp+1n)
(ii) if p=-1, then O(nk loglogn)
5. T(n)= 2T(n/2)+n
(iii) if p<-1, then O(nk)
6. T(n)= 4T(n/2)+n2logn
Case 3: if logba < k, and
(i) if p>=0, then O(nk logpn)
7. T(n)= 2T(n/2)+ n/logn
(ii) if p<0, then O(nk)
8. T(n)= 2T(n/2)+ n/log2n

9. T(n)= T(n/2)+n2
 Recurrence Relation
Use Masters Theorem and find time complexity
1. T(n)= 2T(n/2)+1 T(n) = aT(n/b) +f(n), where a>=1, b>1, and
Here a=2, b=2, f(n)= O(n0 log0n) f(n)= O(nk logpn), where k>=0
So, k=0, p=0
Case 1: if logba > k, then O(nlogba)
Log22=1 and k=0, case 1 is satisfied
T(n)= O(nlog22)= O(n1) Case 2: if logba = k, and
(i) if p>-1, then O(nk logp+1n)
2. T(n)= 4T(n/2)+n
(ii) if p=-1, then O(nk loglogn)
Here a=4, b=2, f(n)= O(n1 log0n)
So, k=1, p=0
(iii) if p<-1, then O(nk)

Log24=2 and k=1, case 1 is satisfied Case 3: if logba < k, and

T(n)= O(nlog24)= O(n2) (i) if p>=0, then O(nk logpn)
(ii) if p<0, then O(nk)
3. T(n)= 8T(n/2)+n2
Here a=8, b=2, f(n)= O(n2 log0n)
So, k=2, p=0, Log28=3 and k=2, case 1 is satisfied
T(n)= O(nlog28)= O(n3)
 Recurrence Relation
Use Masters Theorem and find time complexity
4. T(n)= 8T(n/2)+ n logn
T(n) = aT(n/b) +f(n), where a>=1, b>1, and f(n)= O(nk
Here a=8, b=2, f(n)= O(n logn)
logpn), where k>=0
So, k=1, p=1
log
Case 1: if logba > k, then O(n ba)
Log28=3 and k=1, case 1 is satisfied,
T(n)= O(nlog28)= O(n3) Case 2: if logba = k, and
(i) if p>-1, then O(nk logp+1n)
5. T(n)= 2T(n/2)+n (ii) if p=-1, then O(nk loglogn)
Here a=2, b=2, f(n)= O(n1 log0n) (iii) if p<-1, then O(nk)
So, k=1, p=0
Case 3: if logba < k, and
Log22=1 and k=1, case 2(i) is satisfied (i) if p>=0, then O(nk logpn)
T(n)= O(n1 log0+1n)= O(nlogn) (ii) if p<0, then O(nk)
6. T(n)= 4T(n/2)+n2logn
Here a=4, b=2, f(n)= O(n2 log0n)
So, k=2, p=0
Log24=2 and k=2, case 2(i) is satisfied
T(n)= O(n2 log1+1n)= O(n2 log2 n)
 Recurrence Relation
Use Masters Theorem and find time complexity
7. T(n)= 2T(n/2)+ n/logn
T(n) = aT(n/b) +f(n), where a>=1, b>1, and f(n)= O(nk
Here a=2, b=2, f(n)= O(n/logn)
logpn), where k>=0
So, k=1, p=-1
log
Case 1: if logba > k, then O(n ba)
Log22=1 and k=1, case 2(ii) is satisfied,
T(n)= O(n1loglogn)= O(nloglogn) Case 2: if logba = k, and
(i) if p>-1, then O(nk logp+1n)
8. T(n)= 2T(n/2)+ n/log2n (ii) if p=-1, then O(nk loglogn)
Here a=2, b=2, f(n)= O(n/log2n) (iii) if p<-1, then O(nk)
So, k=1, p=-2
Case 3: if logba < k, and
Log22=1 and k=1, case 2(iii) is satisfied (i) if p>=0, then O(nk logpn)
T(n)= O(n1)= O(n) (ii) if p<0, then O(nk)
9. T(n)= T(n/2)+n2
Here a=1, b=2, f(n)= O(n2 log0n )
So, k=2, p=0
Log21=0 and k=2, case 3(i) is satisfied
T(n)= O(n2 log0n)= O(n2 )
Recurrence Relation
Find the time complexity for the root functions

T(√n)+1 if n>2
T(n)= 1 n=2
Recurrence Relation
T(√n)+1 if n>2
T(n)= 1 n=2

T(n) = T(n1/2 )+1 (i)

= [T(n1/2*2 )+1]+1
T(n) = T(n1/2^2)+2 (ii)
T(n) = T(n1/2^3 )+3 (iii)
.
.
T(n) = T(n1/2^k )+k (iv)
Assume n=2m

T(2m) = T(nm/2^k )+k

Assume T(nm/2^k )=T(2)
m/2^k =1, k= log2m
As n=2m
m= log2n
k= loglog2n
O(loglog2n)