Training Latent Diffusion Models with Interacting Particle Algorithms

Tim Y. J. Wang, Juan Kuntz, O. Deniz Akyildiz
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:2791-2799, 2026.

Abstract

We introduce a novel particle-based algorithm for end-to-end training of latent diffusion models. We reformulate the training task as minimizing a free energy functional and obtain a gradient flow that does so. By approximating the latter with a system of interacting particles, we obtain the algorithm, which we underpin theoretically by providing error guarantees. The novel algorithm compares favorably in experiments with previous particle-based methods and variational inference analogues.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-wang26f, title = { Training Latent Diffusion Models with Interacting Particle Algorithms }, author = {Wang, Tim Y. J. and Kuntz, Juan and Akyildiz, O. Deniz}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {2791--2799}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/wang26f/wang26f.pdf}, url = {https://proceedings.mlr.press/v300/wang26f.html}, abstract = { We introduce a novel particle-based algorithm for end-to-end training of latent diffusion models. We reformulate the training task as minimizing a free energy functional and obtain a gradient flow that does so. By approximating the latter with a system of interacting particles, we obtain the algorithm, which we underpin theoretically by providing error guarantees. The novel algorithm compares favorably in experiments with previous particle-based methods and variational inference analogues. } }
Endnote
%0 Conference Paper %T Training Latent Diffusion Models with Interacting Particle Algorithms %A Tim Y. J. Wang %A Juan Kuntz %A O. Deniz Akyildiz %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-wang26f %I PMLR %P 2791--2799 %U https://proceedings.mlr.press/v300/wang26f.html %V 300 %X We introduce a novel particle-based algorithm for end-to-end training of latent diffusion models. We reformulate the training task as minimizing a free energy functional and obtain a gradient flow that does so. By approximating the latter with a system of interacting particles, we obtain the algorithm, which we underpin theoretically by providing error guarantees. The novel algorithm compares favorably in experiments with previous particle-based methods and variational inference analogues.
APA
Wang, T.Y.J., Kuntz, J. & Akyildiz, O.D.. (2026). Training Latent Diffusion Models with Interacting Particle Algorithms . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:2791-2799 Available from https://proceedings.mlr.press/v300/wang26f.html.

Related Material