Towards a Probabilistic Programming Language for Modelling Complex Systems

The Leeds Institute for Data Analytics is pleased to present the next seminar in our series showcasing data analytics.

Abstract

The keynote speaker will be Daniel Tang, Improbable Worlds Ltd.

Computational models are powerful tools for the analysis of complex systems. However, when the dimension of a model’s state space becomes large it often becomes difficult to calibrate, to update its state in response to new observations and to account properly for uncertainty. This has created a barrier to the wider use of large-scale models of complex systems in real world applications.

Probabilistic programming provides a natural and powerful framework, based on Bayesian inference, within which we can tackle the problems of data assimilation, calibration and uncertainty quantification in computational models. Many existing tools for Bayesian inference do not scale well, but by combining recent advances in machine learning with techniques developed in numerical weather prediction we are building a set of tools that will allow us to scale up inference techniques and apply them to the problems of modelling very large, complex systems.

In this seminar I will demonstrate Keanu, our open-source probabilistic programming language, and show how we’ve used it to model the service quality of a national telecoms network and model the power distribution requirements of a transition to electric vehicles.

 Agenda

15:30-16:00: Speaker to be announced

16:00-17:00: Towards a probabilistic programming language for modelling complex systems– Daniel Tang, Improbable Worlds

17:00-18:00: Networking reception with drinks and nibbles 

This seminar is free and open to all but places must be registered in advance. To book please email Hayley Irving with your name, occupation and faculty/organisation.