nicolasJouvin/MMPCA: Inference and Clustering for Mixture of Multinomial Principal Component Analysis

Cluster any count data matrix with a fixed number of variables, such as document/term matrices. It integrates the dimension reduction aspect of topic models in the mixture models framework. Inference is done by means of a greedy Classification Variational Expectation Maximisation (C-VEM) algorithm. An Integrated Classication Likelihood (ICL) model selection is designed for selecting the latent dimension (number of topics) and the number of clusters. For more details, see the article of Jouvin et. al. (2020) <arxiv:1909.00721>.

install.packages("remotes")
remotes::install_github("nicolasJouvin/MMPCA")