Consistency Training with Physical Constraints
Abstract
We propose a physics-aware Consistency Training (CT) method that accelerates sampling in Diffusion Models with physical constraints. Our approach leverages a two-stage strategy: (1) learning the noise-to-data mapping via CT, and (2) incorporating physics constraints as a regularizer. Experiments on toy examples show that our method generates samples in a single step while adhering to the imposed constraints. This approach has the potential to efficiently solve partial differential equations (PDEs) using deep generative modeling.
Cite
Text
Chang et al. "Consistency Training with Physical Constraints." ICLR 2025 Workshops: FPI, 2025.Markdown
[Chang et al. "Consistency Training with Physical Constraints." ICLR 2025 Workshops: FPI, 2025.](https://mlanthology.org/iclrw/2025/chang2025iclrw-consistency/)BibTeX
@inproceedings{chang2025iclrw-consistency,
title = {{Consistency Training with Physical Constraints}},
author = {Chang, Che-Chia and Dai, Chen-Yang and Lin, Te-Sheng and Lai, Ming-Chih and Lai, Chieh-Hsin},
booktitle = {ICLR 2025 Workshops: FPI},
year = {2025},
url = {https://mlanthology.org/iclrw/2025/chang2025iclrw-consistency/}
}