Additive Non-negative Matrix Factorization for Missing Data

TL;DR

This paper introduces an additive NMF approach that effectively imputes missing data in multivariate datasets, with proven convergence and demonstrated improvements in classification tasks.

Contribution

It presents a novel interpretation of NMF for missing data imputation, along with a joint optimization scheme and convergence proof.

Findings

Effective missing data imputation demonstrated

Convergence of the proposed algorithms proved

Improved classification accuracy with missing attributes

Abstract

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.