Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

YouTube

Equivariant Flow Matching for Deep Generative Models in Physics

Valence Labs via YouTube

Overview

Explore the cutting-edge research on equivariant flow matching in this comprehensive 1-hour 8-minute talk by Leon Klein from Valence Labs. Delve into the world of normalizing flows and their applications in physics, particularly for modeling probability distributions. Learn about Boltzmann generators and their role in tackling the sampling problem in statistical physics. Discover the importance of incorporating symmetries into models for many-body systems like molecules and proteins. Examine the concept of equivariant continuous normalizing flows (CNFs) and understand the challenges in their practical application. Gain insights into the novel equivariant flow matching approach, which offers efficient, simulation-free training of equivariant CNFs. Follow along as Klein demonstrates the effectiveness of this method on many-particle systems and alanine dipeptide, showcasing improved sampling efficiency and scalability. The talk covers key topics such as motivation for the sampling problem, continuous normalizing flows, equivariant Boltzmann generators, conditional flow matching, and equivariant optimal transport flow matching, concluding with results, a summary, and a Q&A session.

Syllabus

- Intro
- Motivation: the sampling problem
- Continuous normalizing flows
- Equivariant boltzmann generator
- Conditional flow matching
- Equivariant flow matching
- Equivariant optimal transport flow matching
- Results
- Summary
- Q+A

Taught by

Valence Labs

Reviews

Start your review of Equivariant Flow Matching for Deep Generative Models in Physics

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.