Probability-Flow ODE in Infinite-Dimensional Function Spaces
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.
Cite
Text
Na et al. "Probability-Flow ODE in Infinite-Dimensional Function Spaces." ICLR 2025 Workshops: DeLTa, 2025.Markdown
[Na et al. "Probability-Flow ODE in Infinite-Dimensional Function Spaces." ICLR 2025 Workshops: DeLTa, 2025.](https://mlanthology.org/iclrw/2025/na2025iclrw-probabilityflow/)BibTeX
@inproceedings{na2025iclrw-probabilityflow,
title = {{Probability-Flow ODE in Infinite-Dimensional Function Spaces}},
author = {Na, Kunwoo and Lee, Junghyun and Yun, Se-Young and Lim, Sungbin},
booktitle = {ICLR 2025 Workshops: DeLTa},
year = {2025},
url = {https://mlanthology.org/iclrw/2025/na2025iclrw-probabilityflow/}
}