LI2022: Logic of Probabilistic Programming, Jan. 31, 2022 – Feb. 4, 2022

Probabilities play an increasing role in computer science. Important deterministically-intractable problems admit feasible approximate probabilistic solutions. Probabilistic algorithms are essential in distributed computing and numerical computing (Monte Carlo methods), not to mention the role of probabilistic methods in deep learning (stochastic gradient). Linear logic shows quite useful in this setting because some of the most basic notions of probability theory are linear: Markov chains or kernels, Bayesian networks etc.

The general purpose of this week is not only to present the state of the art on the applications of proof theory, category theory and denotational semantics to the analysis of probabilistic programs and to the foundation of probabilistic formal methods for program and system verification, but also to stress the specificities of Bayesian programming and machine learning where programs represent statistical models.

The program of the week is designed for conveying to the non-specialist a coherent picture of this exciting and very active area.