Johannes Borgström: Bayesian Probabilistic Programming

Probabilistic programming is a convenient way of specifying generative models for observed data. Starting from a simple probabilistic language that suffices for simple models, I will discuss the addition of different language features for probabilistically sized data, distributions over distributions, model comparison and averaging, and reasoning about other agents' knowledge.

The talk will be at an elementary level, knowledge in basic probability theory (Bayes’ law) will be handy.