Bayesian nonparametrics is the branch of Bayesian analysis in which the prior is specified not in terms of a distribution with a fixed set of parameters, but rather via a stochastic process—an infinite collection of random variables.  The infinite nature of nonparametric models creates both opportunities and challenges.  The opportunities are those of flexibility and robustness—nonparametric approaches supply open-ended sets of degrees of freedom to models, allowing new phenomena to be captured as data accrue, and they often require weaker a priori assumptions than classical parametric models.  On the other hand, the challenges are those of exerting statistical control on the degrees of freedom so that models find signal rather than noise and those of finding effective computational procedures for manipulating stochastic processes under operations of conditioning and marginalization.
Much of the recent growth in interest in Bayesian nonparametrics is driven by the needs of applications.  In particular, data in emerging domains such as document modeling, social network analysis, image processing and natural language processing present many of the features that Bayesian nonparametrics aims to capture.  Accordingly, our development of the field will be application-centric, with models motivated by real-world problems.
The school will make use of lectures, practical sessions, software demonstrations, informal discussion sessions and presentations of research projects by school participants. The slides and background reading material will be distributed to the students before the start of the course.

Scientific coordinators

  • Guido Consonni
  • Fabrizio Ruggeri


Professor Michael Jordan,  Department of Electrical Engineering and Computer Science and  Department of Statistics at the University of California, Berkeley, USA.

Professor Francois Caron, Department of Statistics, University College Oxford, UK.