EVENT: Annual LMS/BCS-FACS Evening Seminar
SPEAKER: Sam Staton, University of Oxford
TITLE: Programming-based foundations for statistics
DATE: Thursday, 17 November 2022, Starting time: 18:00 UTC
VENUE: Online via Zoom
EVENT PAGE: https://www.lms.ac.uk/events/lectures/lms-bcs-facs-evening-seminars
REGISTRATION LINK: https://www.lms.ac.uk/civicrm/event/register?id=88&reset=1
Prior registration is required for attendance -- the registration site
will close on Wednesday, 16 November, at 17:00
Probabilistic programming is a popular tool for statistics and
machine learning. The idea is to describe a statistical model as
a program with random choices. The program might be a simulation
of a system, such as a physics model, a model of viral spread,
or a model of electoral behaviour. We can now carry out
statistical inference over the system, for example, by running a
Monte Carlo simulation – running the simulation 100,000’s of
As I will discuss in this talk, the idea of treating statistical
models as computer programs also has a foundational appeal. If
we can understand statistical models as programs, then the
foundations of probability and statistics can be discussed in
terms of program semantics. There is a chance of new
foundational perspectives on statistics, in terms of programming
languages and their formal methods.
As I will explain, this programming-based foundation for
statistics is attractive because there are some intuitively
simple scenarios, such as inference over function spaces, which
have an easy programming implementation, but for which the
traditional mathematical interpretation is complicated.
Sam Staton is a Professor of Computer Science and Royal Society
University Research Fellow at the University of Oxford. There he
currently runs an ERC grant "Better Languages for Statistics".
Before arriving in Oxford in 2015, Sam spent time in Nijmegen,
Paris, and Cambridge. His PhD was in Cambridge with Marcelo
Fiore (2007). Sam's main research is in programming language
theory, but he is also interested in logic and category theory.
He has recent contributions in probabilistic programming
languages, and quantum computing and programming languages.