A1419
Title: Assessing consistency in Poisson NMF estimation for topic models
Authors: Louisa Kontoghiorghes - Kings College London (United Kingdom) [presenting]
Abstract: Poisson Non-Negative Matrix Factorization (Poisson NMF) is a probabilistic Matrix Factorization method used for modeling count data to uncover latent structures for applications such as text analysis and recommendation systems. The model uncovers latent topics in text corpora by estimating the document-topic and topic-term matrices. The convergence behavior of the model parameters is illustrated through Monte Carlo simulations.
