Bayesian nonparametrics with prior information

Model-based clustering methods typically rely on discrete mixture models, to represent data clusters as arising from different probability distributions. Discrete nonparametric priors, such as Dirichlet and Pitman-Yor processes, are often used in Bayesian settings to jointly infer the number of clusters as well as their characterization. I have been researching on how to embed these priors in large hierarchical models when dealing with complex data structures or substantial prior information. I have been researching on how to embed these priors in large hierarchical models when dealing with complex data structures or substantial prior information.