Cassandra Durr: Discovering Governing Equations using Dynamical Symbolic Regression
Cassandra Durr is a PhD candidate at the STOR-i Centre for Doctoral Training at Lancaster University, UK. Her research focuses on non-linear system identification through equation discovery of physical systems, with an emphasis on epidemic modelling. Cassandra holds a bachelor’s degree in Actuarial Sciences and postgraduate degrees in Mathematical Statistics, and Machine Learning & Artificial Intelligence.
Time and place:–,
Erling Sverdrups plass, Niels Henrik Abels hus, 8th floor
In this talk, we address the problem of dynamical system identification for ordinary differential equation systems where the underlying system structure (or stoichiometry) is known, but the functional form of the flux equations are not. We propose a grey-box method that uses dynamical symbolic regression to infer these functional forms directly from observational data. We demonstrate that this approach accurately recovers interpretable mathematical expressions from noisy data across a range of non-linear dynamical systems.
