It is primarily intended to make life easier for cognitive science and machine learning researchers who use Bayesian nonparametric models in their work.
It is a collection of OpenCL-accelerated samplers that perform Bayesian posterior inferences for Bayesian nonparametric models (such as the Chinese Restaurant Process mixture models and the Indian Buffet Process models). MPBNP is short for Massively Parallel Bayesian Non parametric models.
