Recovering Hidden Diagonal Structures via Non-Negative Matrix Factorization with Multiple Constraints

摘要: Revealing data with intrinsically diagonal block structures is particularly useful for analyzing groups of highly correlated variables. Earlier researches based on non-negative matrix factorization (NMF) have been shown to be effective in representing such by decomposing the observed into two factors, where one factor considered feature and other expansion loading from a linear algebra perspective. If are sampled multiple independent subspaces, would possess structure under an ideal decomposition. However, standard NMF method its variants not reported exploit this type via direct estimation. To address issue, constraints model proposed paper. The include sparsity norm total variational each column matrix. capable efficiently recovering hidden samples. An efficient numerical algorithm using alternating direction multipliers optimizing new model. Compared several benchmark models, performs robustly effectively simulated real biological data.