English

Probability-Flow ODE in Infinite-Dimensional Function Spaces

Machine Learning 2025-03-14 v1 Machine Learning

Abstract

Recent advances in infinite-dimensional diffusion models have demonstrated their effectiveness and scalability in function generation tasks where the underlying structure is inherently infinite-dimensional. To accelerate inference in such models, we derive, for the first time, an analog of the probability-flow ODE (PF-ODE) in infinite-dimensional function spaces. Leveraging this newly formulated PF-ODE, we reduce the number of function evaluations while maintaining sample quality in function generation tasks, including applications to PDEs.

Keywords

Cite

@article{arxiv.2503.10219,
  title  = {Probability-Flow ODE in Infinite-Dimensional Function Spaces},
  author = {Kunwoo Na and Junghyun Lee and Se-Young Yun and Sungbin Lim},
  journal= {arXiv preprint arXiv:2503.10219},
  year   = {2025}
}

Comments

26 pages, 8 figures. Accepted to the ICLR 2025 DeLTa Workshop