A blind sparse approach for estimating constraint matrices in Paralind data models

摘要: In this paper we address the problem of estimating interaction matrices Paralind decomposition. general, is an ill-posed admitting infinite number solutions. First study gain imposing sparsity constraints on matrices, in terms model identifiability. Then, propose a new algorithm (S-Paralind) for fitting model, using l 2 — 1 optimization step matrix. This approach provides more accurate and robust estimates constraint than ALS-Paralind, thus improving interpretabil-ity