Nonnegative Matrix Factorization with KL-Divergnece

This program is a MATLAB implementation of the NMF on the face data.

The function 'nmf' does the original NMF algorithm with the KL-Divergence (shown below) minimization, and the function 'ssnmf' does the sparsity imposed NMF.

KL-Divergence:

Matrix update is given by:

Where B is the factorized matrix with a user defined rank and W is the factorized weighting matrix.

An example result: