25 Years of Dependent Dirichlet Processes
Sunday, June 22nd, 2025
9:00am – 12:30pm
The Dependent Dirichlet Process (DDP) was first introduced by a seminal Proceedings article and a technical report by Steven MacEachern in 1999/2000; this pioneering work extended the classical Dirichlet Process, introducing a framework for modeling dependencies across multiple distributions and enabling flexible modeling of complex data structures. To mark the 25th anniversary of the DDP, we are organizing a workshop that will bring together researchers reflecting on the profound impact of MacEachern’s technical report and its contributions to theory, methods, and applications. This event will not only celebrate Steven MacEachern’s groundbreaking contribution but also honor the vibrant community of researchers who have expanded on the original ideas.
Registration is mandatory and it is included for free when registering to the BNP 14 Conference
Session 1 (8:30 – 10:15)
8:30–9:00: Felipe Barrientos, Florida State University, USA
On MacEachern’s Dependent Dirichlet Process: Flexible Bayesian Predictor-Dependent Modeling
9:00–9:30: Jim Griffin, University College London, UK
DDP modelling using normalised random measures
9:30–10:00: Antonio Lijoi, Bocconi University, Italy
The DDP and random measures-based approaches
10:00–10:15: A Perspective: Maria De Iorio, National University of Singapore, Singapore
Coffee Break (10:15 – 10:30)
Session 2 (10:30–12:15)
10:30–11:00: Ramses Mena, Universidad Nacional Autónoma de México (UNAM), Mexico
Continuous time Bayesian nonparametric dependent processes
11:00–11:30: David Dunson, Duke University, USA
Logistic-beta processes for dependent random probabilities with beta marginals
11:30–12:00: Tamara Broderick, Massachusetts Institute of Technology (MIT), USA
The dependent Dirichlet process and local exchangeability
12:00–12:15: Final Perspective: Steve MacEachern, Ohio State University, USA
Q&A/Wrap‑up (12:15–12:30)
Stay tuned for updates, and we look forward to seeing you at UCLA!
Contact the organizers, Alejandro Jara (ajara AT mat.puc.cl ) and Michele Guindani (mguindani AT ucla.edu) for additional information.
Workshop on Bayesian Predictive Inference
Sunday June 22nd, 2025
2:30pm – 6:30pm
The “Workshop on Bayesian Predictive Inference“ is a satellite workshop to BNP14, set to take place on the afternoon of June 22nd, 2025 atUCLA in Los Angeles, United States.
Within the field of Bayesian statistics, there has been a recent surge in the development of methods that bring the focus onto the predictive distribution. These methods can bypass the usual prior-likelihood construction by directly considering the predictive rule and can recover inferential uncertainty through the imputation of missing data. The predictive approach provides an intriguing alternative to the usual prior-likelihood focus of Bayesian models, promising new computational, theoretical, and foundational opportunities.
This workshop aims to bring together researchers and experts to present and discuss recent advances in predictive methods for Bayesian inference. The event will feature a blend of tutorials and talks.
Tutorials
2:30–3:30: Sandra Fortini, Bocconi University, Italy
On the theoretical foundations of predictive Bayesian learning
3:30–4:30: Lorenzo Cappello, Universitat Pompeu Fabra, Spain
Recursive methods for predictive Bayes
Coffee Break (4:30 – 4:50)
Research Talks
4:50–6:10: Bernardo Flores López, University of Texas, Austin, USA
6:10–6:30: Carlos Erwin Rodríguez, IIMAS-UNAM, Mexico
Martingale Posterior Inference for Finite Mixture Models and Clustering
6:30–6:50: Hristo Sariev, Institute of Mathematics and Informatics, Bulgarian Academy of Sciences, Bulgaria
Characterization of exchangeable Hoeffding decomposable sequences
6:50–7:10: Shengjun (Percy) Zhai, The University of Chicago, USA
Minimax Bayesian Predictive Inference with the Horseshoe Prior
Q&A (7:10–7:20)
Contact the organizers, Edwin Fong (chefong AT hku.hk ), Sonia Petrone (Sonia. Petrone AT unibocconi.it), and Stephen Walker (s.g.walker AT math.utexas.edu) for additional information.