Diffusion Rejection Sampling

Byeonghu Na, Yeongmin Kim, Minsang Park, Donghyeok Shin, Wanmo Kang, Il-Chul Moon
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:37097-37121, 2024.

Abstract

Recent advances in powerful pre-trained diffusion models encourage the development of methods to improve the sampling performance under well-trained diffusion models. This paper introduces Diffusion Rejection Sampling (DiffRS), which uses a rejection sampling scheme that aligns the sampling transition kernels with the true ones at each timestep. The proposed method can be viewed as a mechanism that evaluates the quality of samples at each intermediate timestep and refines them with varying effort depending on the sample. Theoretical analysis shows that DiffRS can achieve a tighter bound on sampling error compared to pre-trained models. Empirical results demonstrate the state-of-the-art performance of DiffRS on the benchmark datasets and the effectiveness of DiffRS for fast diffusion samplers and large-scale text-to-image diffusion models. Our code is available at https://github.com/aailabkaist/DiffRS.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-na24a, title = {Diffusion Rejection Sampling}, author = {Na, Byeonghu and Kim, Yeongmin and Park, Minsang and Shin, Donghyeok and Kang, Wanmo and Moon, Il-Chul}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {37097--37121}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/na24a/na24a.pdf}, url = {https://proceedings.mlr.press/v235/na24a.html}, abstract = {Recent advances in powerful pre-trained diffusion models encourage the development of methods to improve the sampling performance under well-trained diffusion models. This paper introduces Diffusion Rejection Sampling (DiffRS), which uses a rejection sampling scheme that aligns the sampling transition kernels with the true ones at each timestep. The proposed method can be viewed as a mechanism that evaluates the quality of samples at each intermediate timestep and refines them with varying effort depending on the sample. Theoretical analysis shows that DiffRS can achieve a tighter bound on sampling error compared to pre-trained models. Empirical results demonstrate the state-of-the-art performance of DiffRS on the benchmark datasets and the effectiveness of DiffRS for fast diffusion samplers and large-scale text-to-image diffusion models. Our code is available at https://github.com/aailabkaist/DiffRS.} }
Endnote
%0 Conference Paper %T Diffusion Rejection Sampling %A Byeonghu Na %A Yeongmin Kim %A Minsang Park %A Donghyeok Shin %A Wanmo Kang %A Il-Chul Moon %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-na24a %I PMLR %P 37097--37121 %U https://proceedings.mlr.press/v235/na24a.html %V 235 %X Recent advances in powerful pre-trained diffusion models encourage the development of methods to improve the sampling performance under well-trained diffusion models. This paper introduces Diffusion Rejection Sampling (DiffRS), which uses a rejection sampling scheme that aligns the sampling transition kernels with the true ones at each timestep. The proposed method can be viewed as a mechanism that evaluates the quality of samples at each intermediate timestep and refines them with varying effort depending on the sample. Theoretical analysis shows that DiffRS can achieve a tighter bound on sampling error compared to pre-trained models. Empirical results demonstrate the state-of-the-art performance of DiffRS on the benchmark datasets and the effectiveness of DiffRS for fast diffusion samplers and large-scale text-to-image diffusion models. Our code is available at https://github.com/aailabkaist/DiffRS.
APA
Na, B., Kim, Y., Park, M., Shin, D., Kang, W. & Moon, I.. (2024). Diffusion Rejection Sampling. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:37097-37121 Available from https://proceedings.mlr.press/v235/na24a.html.

Related Material