Speaker |
Jiajun He |
|---|---|
Affiliation |
University of Cambridge |
Date |
Friday, 11 April 2025 |
Time |
12:00-13:00 |
Location |
UCL Centre for Artificial Intelligence, 1st Floor, 90 High Holborn, London WC1V 6BH |
Link |
https://ucl.zoom.us/j/99748820264 |
Event series |
Jump Trading/ELLIS CSML Seminar Series |
Abstract |
Free energy estimation is a fundamental challenge in machine learning (normalization factors) and statistical physics (partition functions) with broad applications in science. In this talk, we describe a new framework for estimating the free energy difference between two systems via a learned non-equilibrium transport. Our approach yields variational upper and lower bounds on free energy differences, alongside a consistent, minimum-variance estimator. Our approach unifies equilibrium averages and non-equilibrium transports under a single framework, and achieves significant improvements over existing neural network methods. |
Biography |
Jiajun He (https://jiajunhe98.github.io/) is a PhD student at the University of Cambridge, supervised by Prof. José Miguel Hernández-Lobato. His research interests include probabilistic models, with a particular interest in their applications to sampling or AI for science. |