Algorithms for Non-Negative Matrix Factorization – Short Science Summary:
This paper analyzes two separate iterative methods for calculating NMF decompositions: one that minimizes least-squares error and one that minimizes generalized KL-divergence.
The cost functions given for the two methods are as follows:
||A – B||^2 = ∑_(ij) (Aij*Bij)^2
D(A||B) = ∑_ij (Aij log(Aij/Bij) - Aij+Bij)
The authors found that using the below multiplicative update rules for each iteration of the corresponding optimization algorithm results in a good compromise between run speed and ease of implementation:
The authors go on to prove convergence to at least a locally optimal solution for both of the above choices of cost function.