Ella Bingham, HUT/CIS
In this talk we consider the problem of finding latent structure in high dimensional data. It is assumed that the observed data are generated by unknown latent variables and their interactions. The task is to find these latent variables and the way they interact, given the observed data only. We often also assume that the latent variables do not depend on each other but act independently.
First, the problem setting is discussed from several different viewpoints, such as clustering, matrix decompositions and Bayesian data analysis. Then, different methods for estimating the latent structure are presented (mixture models, ICA, PLSA, topic models etc.) Practical applications are discussed, too.