Understanding the Probabilistic Latent Component Analysis Framework

Abstract

Probabilistic Component Latent Analysis (PLCA) is a statistical modeling method for feature extraction from non-negative data. It has been fruitfully applied to various research fields of information retrieval. However, the EM-solved optimization problem coming with the parameter estimation of PLCA-based models has never been properly posed and justified. We then propose in this short paper to re-define the theoretical framework of this problem, with the motivation of making it clearer to understand, and more admissible for further developments of PLCA-based computational systems.