Nonnegative rank factorization -- a heuristic approach via rank reduction (Q2248962)

Given the class of nonnegative \(m \times n\) matrices \(A\), with \(\mathrm{rank}(A)=\mathrm{rank}_+(A)=k\), where \(\mathrm{rank}_+(A)\) is the nonnegative rank of \(A\), the authors consider nonnegative rank factorizations (NRF), that is, factorizations \(A=UV\), where \(U\) and \(V\) are of dimensions \(m \times k\) and \(k \times n\), respectively. An algorithm for the NRF with a respective MATLAB code for such factorization is provided. Furthermore, some examples illustrate the presented method.