The BNP Workshop series brings together the international community of researchers working on Bayesian nonparametric methods. These workshops have been instrumental in shaping the development of the field, providing a focused forum where new theoretical results, computational methods, and applications of nonparametric Bayesian models are presented and debated.
History and Evolution
The BNP Workshop series began as the field of Bayesian nonparametrics was transitioning from a largely theoretical discipline to one with a rapidly expanding range of practical applications. The workshops have been held biennially at locations around the world, reflecting the international character of the BNP research community.
Early workshops bring together the small but growing community of BNP researchers, establishing the series as the field's defining conference.
The workshop grows as BNP methods gain prominence in machine learning and applied statistics, with meetings in locations including Jeju (South Korea), Cambridge (UK), and Melbourne (Australia).
The BNP Workshop attracts researchers from statistics, machine learning, and computer science, reflecting the cross-disciplinary impact of nonparametric Bayesian methods.
The workshop continues to evolve, incorporating new topics such as deep generative models, Bayesian nonparametric methods for causal inference, and scalable BNP computation.
The BNP Workshop is one of the few conferences that genuinely bridges the statistics and machine learning communities. Topics such as Dirichlet process mixtures, Gaussian processes, Indian buffet processes, and Bayesian neural networks attract researchers from both fields, creating a productive intellectual exchange that has enriched both disciplines.
Scientific Scope
The workshop covers the full range of Bayesian nonparametric research. Key topics include:
Random probability measures: Dirichlet processes, Pitman-Yor processes, normalized random measures, and their extensions. Random functions: Gaussian processes, deep Gaussian processes, and other priors on function spaces. Random structures: Indian buffet processes, random graphs, and models for combinatorial objects. Computational methods: MCMC for BNP models, variational inference for infinite-dimensional models, and scalable approximations. Applications: clustering, density estimation, survival analysis, topic modeling, and beyond.
Community and Atmosphere
The BNP Workshop is known for its collegial and intellectually stimulating atmosphere. With typically 150-250 participants, it is small enough to foster genuine interaction and discussion, yet large enough to represent the full breadth of the field. Extended discussion periods, poster sessions, and social events create opportunities for the kind of deep engagement that is often difficult at larger conferences.
"The BNP Workshop is where the hardest and most exciting problems in nonparametric Bayesian statistics are tackled. It is the heartbeat of the BNP community."— BNP researcher