Bayesian non-parametric parsimonious clustering

This paper proposes a new Bayesian non-parametric approach for clustering. It relies on an infinite Gaussian mixture model with a Chi- nese Restaurant Process (CRP) prior, and an eigenvalue decomposition of the covariance matrix of each cluster. The CRP prior allows to control the model complexity in a principled way and to automatically learn the number of clusters. The covariance matrix decomposition allows to fit var- ious parsimonious models going from simplest spherical ones to the more complex general one. We develop an MCMC Gibbs sampler to learn the models. First results obtained on both simulated and real data highlight the interest of the proposed infinite parsimonious mixture model.