Additive Non-negative Matrix Factorization for Missing Data
It addresses the problem of missing data in multivariate analysis for practitioners using NMF, but the contribution appears incremental.
The paper proposes an additive NMF approach to handle missing data by jointly optimizing missing attributes and NMF factors, proving monotonic convergence. Classification results are presented for missing attribute scenarios.
Non-negative matrix factorization (NMF) has previously been shown to be a useful decomposition for multivariate data. We interpret the factorization in a new way and use it to generate missing attributes from test data. We provide a joint optimization scheme for the missing attributes as well as the NMF factors. We prove the monotonic convergence of our algorithms. We present classification results for cases with missing attributes.