pyDNMFk is a software package for applying non-negative matrix factorization in a distributed fashion to large datasets. It has the ability to minimize the difference between reconstructed data and the original data through various norms (Frobenious, KL-divergence).