作者: Andrzej Cichocki , Rafal Zdunek , Shun-ichi Amari
DOI: 10.1007/11679363_5
关键词: Sparse approximation 、 Algorithm 、 Smoothness (probability theory) 、 Non-negative matrix factorization 、 Computer science 、 Source separation 、 Independent component analysis 、 Robustness (computer science) 、 Blind signal separation 、 Principal component analysis 、 Outlier 、 Divergence (statistics) 、 Matrix decomposition 、 Divergence 、 Regularization (mathematics) 、 Generalized function
摘要: In this paper we discus a wide class of loss (cost) functions for non-negative matrix factorization (NMF) and derive several novel algorithms with improved efficiency and robustness to …