Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2502.13280 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908922188660736 |
|---|---|
| author | Hwang, Himchan Jeong, Hyeokju Shin, Dong Kyu Park, Che-Sang Kweon, Sehee Yoon, Sangwoong Park, Frank Chongwoo |
| author_facet | Hwang, Himchan Jeong, Hyeokju Shin, Dong Kyu Park, Che-Sang Kweon, Sehee Yoon, Sangwoong Park, Frank Chongwoo |
| contents | We propose the Value Gradient Sampler (VGS), a diffusion sampler parameterized by value functions. VGS generates samples from an unnormalized target density (i.e., energy) by evolving randomly initialized particles along the gradient of the value function. In many sampling problems where the target density exhibits invariant symmetries, value functions provide a novel approach to leveraging invariant networks for sampling by inducing an equivariant gradient flow, without requiring more complex equivariant networks. The value networks are trained via temporal difference learning, which supports off-policy training and other established reinforcement learning (RL) techniques. By combining advanced RL methods with efficient invariant networks, VGS achieves both the highest sample quality and the fastest sampling speed among our baselines on the 55-particle Lennard-Jones system. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_13280 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Value Gradient Sampler: Learning Invariant Value Functions for Equivariant Diffusion Sampling Hwang, Himchan Jeong, Hyeokju Shin, Dong Kyu Park, Che-Sang Kweon, Sehee Yoon, Sangwoong Park, Frank Chongwoo Machine Learning We propose the Value Gradient Sampler (VGS), a diffusion sampler parameterized by value functions. VGS generates samples from an unnormalized target density (i.e., energy) by evolving randomly initialized particles along the gradient of the value function. In many sampling problems where the target density exhibits invariant symmetries, value functions provide a novel approach to leveraging invariant networks for sampling by inducing an equivariant gradient flow, without requiring more complex equivariant networks. The value networks are trained via temporal difference learning, which supports off-policy training and other established reinforcement learning (RL) techniques. By combining advanced RL methods with efficient invariant networks, VGS achieves both the highest sample quality and the fastest sampling speed among our baselines on the 55-particle Lennard-Jones system. |
| title | Value Gradient Sampler: Learning Invariant Value Functions for Equivariant Diffusion Sampling |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2502.13280 |