Date
October 9 (Wed) at 15:00 - 16:30, 2024 (JST)
Speaker
  • Yuji Hirono (Assistant Professor, Department of Physics, Graduate School of Science, Osaka University)
Language
English
Host
Lingxiao Wang

Diffusion models have emerged as powerful tools in generative modeling, especially in image generation tasks. In this talk, we introduce a novel perspective by formulating diffusion models using the path integral method introduced by Feynman for describing quantum mechanics. We find this formulation providing comprehensive descriptions of score-based diffusion generative models, such as the derivation of backward stochastic differential equations and loss functions for optimization. The formulation accommodates an interpolating parameter connecting stochastic and deterministic sampling schemes, and this parameter can be identified as a counterpart of Planck's constant in quantum physics. This analogy enables us to apply the Wentzel-Kramers-Brillouin (WKB) expansion, a well-established technique in quantum physics, for evaluating the negative log-likelihood to assess the performance disparity between stochastic and deterministic sampling schemes.

Reference

  1. Yuji Hirono, Akinori Tanaka, Kenji Fukushima, Understanding Diffusion Models by Feynman's Path Integral, Proceedings of the 41st International Conference on Machine Learning (ICML), PMLR 235:18324-18351 (2024), arXiv: 2403.11262

This is a closed event for scientists. Non-scientists are not allowed to attend. If you are not a member or related person and would like to attend, please contact us using the inquiry form. Please note that the event organizer or speaker must authorize your request to attend.

Inquire about this event

Related Link