Bayesian nonparametric regression has traditionally been studied as two separate disciplines. On the one hand researchers have considered the construction of probability measures on spaces of univariate density functions, which act as prior distributions with full coverage in Bayesian inference. These prior distributions are typically achieved by the construction of stochastic processes; the most well known being the Dirichlet process. On the other hand, researchers in stochastic modelling have investigated the construction of probability measures on general function spaces which act as regression functions in specifying the evolution of parametric distributions indexed on a covariate set, such as the mean function of a Gaussian process. To date there has been surprisingly little overlap between these two strands of research.
The aim of the programme is to bring together researchers working in both areas of Bayesian nonparametrics in order to stimulate further research and discuss the foundations of fully nonparametric regression, in theory, methods and applications.