335
v1v2v3 (latest)

Partial Identifiability for Nonnegative Matrix Factorization

SIAM Journal on Matrix Analysis and Applications (SIMAX), 2022
Abstract

Given a nonnegative matrix factorization, RR, and a factorization rank, rr, Exact nonnegative matrix factorization (Exact NMF) decomposes RR as the product of two nonnegative matrices, CC and SS with rr columns, such as R=CSR = CS^\top. A central research topic in the literature is the conditions under which such a decomposition is unique/identifiable, up to trivial ambiguities. In this paper, we focus on partial identifiability, that is, the uniqueness of a subset of columns of CC and SS. We start our investigations with the data-based uniqueness (DBU) theorem from the chemometrics literature. The DBU theorem analyzes all feasible solutions of Exact NMF, and relies on sparsity conditions on CC and SS. We provide a mathematically rigorous theorem of a recently published restricted version of the DBU theorem, relying only on simple sparsity and algebraic conditions: it applies to a particular solution of Exact NMF (as opposed to all feasible solutions) and allows us to guarantee the partial uniqueness of a single column of CC or SS. Second, based on a geometric interpretation of the restricted DBU theorem, we obtain a new partial identifiability result. This geometric interpretation also leads us to another partial identifiability result in the case r=3r=3. Third, we show how partial identifiability results can be used sequentially to guarantee the identifiability of more columns of CC and SS. We illustrate these results on several examples, including one from the chemometrics literature.

View on arXiv
Comments on this paper