This is one of the most used techniques in competitive programming. Let us first consider below simple question.

What is the minimum time complexity to find n’th Fibonacci Number?

We can find n’th Fibonacci Number in O(Log n) time using Matrix Exponentiation. Refer method 4 of this for details. In this post, a general implementation of Matrix Exponentiation is discussed.

For solving the matrix exponentiation we are assuming a linear recurrence equation like below: F(n) = a*F(n-1) + b*F(n-2) + c*F(n-3) for n >= 3 . . . . . Equation (1) where a, b and c are constants. For this recurrence relation it depends on three previous values. Now we will try to represent Equation (1) in terms of matrix. [First Matrix] = [Second matrix] * [Third Matrix] | F(n) | = Matrix 'C' * | F(n-1) | | F(n-1) | | F(n-2) | | F(n-2) | | F(n-3) | Dimension of the first matrix is 3 x 1 . Dimension of third matrix is also 3 x 1. So the dimension of the second matrix must be 3 x 3 [For multiplication rule to be satisfied.] Now we need to fill the Matrix 'C'. So according to our equation. F(n) = a*F(n-1) + b*F(n-2) + c*F(n-3) F(n-1) = F(n-1) F(n-2) = F(n-2) C = [a b c 1 0 0 0 1 0] Now the relation between matrix becomes : [First Matrix] [Second matrix] [Third Matrix] | F(n) | = | a b c | * | F(n-1) | | F(n-1) | | 1 0 0 | | F(n-2) | | F(n-2) | | 0 1 0 | | F(n-3) | Lets assume the initial values for this case :- F(0) = 0 F(1) = 1 F(2) = 1 So, we need to get F(n) in terms of these values. So, for n = 3 Equation (1) changes to | F(3) | = | a b c | * | F(2) | | F(2) | | 1 0 0 | | F(1) | | F(1) | | 0 1 0 | | F(0) | Now similarly for n = 4 | F(4) | = | a b c | * | F(3) | | F(3) | | 1 0 0 | | F(2) | | F(2) | | 0 1 0 | | F(1) | - - - - 2 times - - - | F(4) | = | a b c | * | a b c | * | F(2) | | F(3) | | 1 0 0 | | 1 0 0 | | F(1) | | F(2) | | 0 1 0 | | 0 1 0 | | F(0) | So for n, the Equation (1) changes to - - - - - - - - n -2 times - - - - - | F(n) | = | a b c | * | a b c | * ... * | a b c | * | F(2) | | F(n-1) | | 1 0 0 | | 1 0 0 | | 1 0 0 | | F(1) | | F(n-2) | | 0 1 0 | | 0 1 0 | | 0 1 0 | | F(0) | | F(n) | = [ | a b c | ] ^ (n-2) * | F(2) | | F(n-1) | [ | 1 0 0 | ] | F(1) | | F(n-2) | [ | 0 1 0 | ] | F(0) |

So we can simply multiply our Second matrix n-2 times and then multiply it with the third matrix to get the result. Multiplication can be done in (log n) time using Divide and Conquer algorithm for power (See this or this)

Let us consider the problem of finding n’th term of a series defined using below recurrence.

n'th term, F(n) = F(n-1) + F(n-2) + F(n-3), n >= 3 Base Cases : F(0) = 0, F(1) = 1, F(2) = 1

We can find n’th term using following :

Putting a = 1, b = 1 and c = 1 in above formula | F(n) | = [ | 1 1 1 | ] ^ (n-2) * | F(2) | | F(n-1) | [ | 1 0 0 | ] | F(1) | | F(n-2) | [ | 0 1 0 | ] | F(0) |

Below is C++ implementation of above idea.

// C++ program to find value of f(n) where f(n) // is defined as // F(n) = F(n-1) + F(n-2) + F(n-3), n >= 3 // Base Cases : // F(0) = 0, F(1) = 1, F(2) = 1 #include<bits/stdc++.h> using namespace std; // A utility function to multiply two matrices // a[][] and b[][]. Multiplication result is // stored back in b[][] void multiply(int a[3][3], int b[3][3]) { // Creating an auxiliary matrix to store elements // of the multiplication matrix int mul[3][3]; for (int i = 0; i < 3; i++) { for (int j = 0; j < 3; j++) { mul[i][j] = 0; for (int k = 0; k < 3; k++) mul[i][j] += a[i][k]*b[k][j]; } } // storing the muliplication resul in a[][] for (int i=0; i<3; i++) for (int j=0; j<3; j++) a[i][j] = mul[i][j]; // Updating our matrix } // Function to compute F raise to power n-2. int power(int F[3][3], int n) { int M[3][3] = {{1,1,1}, {1,0,0}, {0,1,0}}; // Multiply it with initial values i.e with // F(0) = 0, F(1) = 1, F(2) = 1 if (n==1) return F[0][0] + F[0][1]; power(F, n/2); multiply(F, F); if (n%2 != 0) multiply(F, M); // Multiply it with initial values i.e with // F(0) = 0, F(1) = 1, F(2) = 1 return F[0][0] + F[0][1] ; } // Return n'th term of a series defined using below // recurrence relation. // f(n) is defined as // f(n) = f(n-1) + f(n-2) + f(n-3), n>=3 // Base Cases : // f(0) = 0, f(1) = 1, f(2) = 1 int findNthTerm(int n) { int F[3][3] = {{1,1,1}, {1,0,0}, {0,1,0}} ; return power(F, n-2); } // Driver code int main() { int n = 5; cout << "F(5) is " << findNthTerm(n); return 0; }

Output :

F(5) is 7

Time Complexity of this solution : O(log n)

This article is contributed by **Abhiraj Smit**. Please write comments if you find anything incorrect, or you want to share more information about the topic discussed above