University of Cambridge > Talks.cam > DAMTP Statistical Physics and Soft Matter Seminar > Dissipation and Optimal Transport in Discrete and Continuous Flow-based Models

Dissipation and Optimal Transport in Discrete and Continuous Flow-based Models

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Sarah Loos.

In this talk, I will discuss connections between control problems that arise in nonequilibrium statistical physics and generative models to illustrate how tools built for low dissipation control and rare event sampling can be leveraged to improve sample efficiency in flow-based generative models. Employing these notions, I will outline a strategy for parameterizing discrete diffusion models using tensor networks, which improves MCMC in some simple models from statistical mechanics. Going beyond this approach, I will discuss “mixed-resolution” discrete and continuous models we have been developing and how to build statistically controlled sampling schemes for these models.

This talk is part of the DAMTP Statistical Physics and Soft Matter Seminar series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2025 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity