Understanding and Exploiting the Connections between NMF and SVM

Support Vector Machines (SVM) and Nonnegative Matrix Factorization (NMF) are standard tools for data analysis. We explore the connections between these two problems, thereby enabling us to import algorithms from SVM world to solve NMF and vice-versa. In particular, one such algorithm developed to solve SVM is adapted to solve NMF. Empirical results show that this new algorithm is competitive with the state-of-the-art NMF solvers.