Progressive Tempering Sampler with Diffusion

TL;DR

This paper introduces PTSD, a novel neural sampling method that combines diffusion models and parallel tempering to produce efficient, uncorrelated samples across temperature levels, surpassing existing diffusion-based neural samplers.

Contribution

The paper proposes PTSD, a sequential diffusion model training approach across temperatures that enhances sampling efficiency and sample independence compared to prior neural samplers.

Findings

PTSD outperforms existing diffusion-based neural samplers in target evaluation efficiency.

It produces well-mixed, uncorrelated samples across temperature levels.

The method effectively reuses sample information to improve sampling quality.

Abstract

Recent research has focused on designing neural samplers that amortize the process of sampling from unnormalized densities. However, despite significant advancements, they still fall short of the state-of-the-art MCMC approach, Parallel Tempering (PT), when it comes to the efficiency of target evaluations. On the other hand, unlike a well-trained neural sampler, PT yields only dependent samples and needs to be rerun -- at considerable computational cost -- whenever new samples are required. To address these weaknesses, we propose the Progressive Tempering Sampler with Diffusion (PTSD), which trains diffusion models sequentially across temperatures, leveraging the advantages of PT to improve the training of neural samplers. We also introduce a novel method to combine high-temperature diffusion models to generate approximate lower-temperature samples, which are minimally refined using MCMC and used to train the next diffusion model. PTSD enables efficient reuse of sample information across temperature levels while generating well-mixed, uncorrelated samples. Our method significantly improves target evaluation efficiency, outperforming diffusion-based neural samplers.