Ohad's Research - Foundations for Type-Driven Probabilistic Modelling [Laboratory for Software Science Tartu Course]

Foundations for type-driven probabilistic modelling

Laboratory for Software Science Tartu Supernumerary Course

Ohad Kammar January 15 2024

The last few years have seen several breakthroughs in the semantic foundations of probabilistic and statistical modelling. In this tutorial, we will use types to introduce, use, and organise abstractions for probabilistic modelling.

We will do so first for discrete probability, and then more generally with the recently-developed quasi-Borel spaces. The course is accompanied by exercises for self-study, allowing you to develop a working knowledge and hands-on experience after it.

Course materials

Lectures

All lectures.

Planned structure:

  1. Lecture 1
  • Motivation
  • Language of probability and distribution
  • Discrete model
  • Simply-typed probability
  • Dependently-typed probability
  1. Lecture 2
  • events, Borel sets, and measurable spaces
  • quasi-Borel spaces
  • Simple type structure
  • Dependent-type structure
  • Standard Borel spaces
  • Integration
  • Random variables
  • Conditioning
Relevant exercise sheets: