Gaussian Mixture Impact Driven by PLSA Models

We present a new topic model that simultaneously fits a sequence of discrete-time data and a corpus of documents with their corresponding publication date. A mixture of normal distributions will be used to model the time-series, while a PLSA approach will be followed for the corpus of documents. The parameters of both models will be coupled, and hence will influence each other.

Phd supervisor: Julio Gonzalo Arroyo