Authors : Bastien Billot and Rémy Deshayes
This is a very short project building on the 2006 paper A Predictive View of Bayesian Clustering by Fernando A. Quintana.
In his paper, Quintana reviews various probabilistic methods for set partitioning, and their interactions. In other words, Quintana is interested in models given as conditional probabilities of either joining an already existing cluster or forming a new one.
Firstly, we recall the paper theoretical framework. Then, we review Quintana's implementations and results - notably, the various Model-Based Clustering (MBC) initializations.
Secondly, we move away from the paper per se and introduce two MBC implementations before comparing them with traditional clustering methods.