Data Free Training of Diffusion Models for Lattice Field Theories

Key points

• Diffusion models are promising for this purpose, but their standard denoising score matching objective requires samples from the target distribution, seemingly reintroducing the original sampling problem.1
• Here, we show that diffusion models can instead be trained from their own generated samples using the target score identity, provided access to the unnormalized probability density.1
• Importance-weighted formulations of this identity yield a consistent target throughout training, while unweighted variants define a self-consistent fixed-point training scheme.1
• We benchmark the different approaches in two-dimensional \(φ^4\) theory and identify effective choices that enable efficient training and sampling across the symmetric, critical, and broken phases.1

Sources (1)

• [1]Data Free Training of Diffusion Models for Lattice Field Theories
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 05:59 PM

  Sampling equilibrium field configurations remains a major computational bottleneck in lattice field theory, motivating the development of generative samplers that can produce independent configurations.

  Diffusion models are promising for this purpose, but their standard denoising score matching objective requires samples from the target distribution, seemingly reintroducing the original sampling problem.

Extractive summary: sentences quoted from the sources.