A0433
Title: Bayesian clustering of high-dimensional data via latent repulsive mixtures
Authors: Alessandra Guglielmi - Politecnico di Milano (Italy) [presenting]
Abstract: Model-based clustering of moderate or large dimensional data is notoriously difficult. A model is proposed for simultaneous dimensionality reduction and clustering by assuming a mixture model for a set of latent scores, which are then linked to the observations via a Gaussian latent factor model. Other researchers recently investigated this approach. A factor-analytic representation is used, and a mixture model is assumed for the latent factors. However, performance can deteriorate in the presence of model misspecification. Assuming a repulsive point process prior to the component-specific means of the mixture for the latent scores is shown to yield a more robust model that outperforms the standard mixture model for the latent factors in several simulated scenarios. To favour well-separated clusters of data, the repulsive point process must be anisotropic, and its density should be tractable for efficient posterior inference. These issues are addressed by proposing a general construction for anisotropic determinantal point processes.
