Implicit Diffusion: Efficient Optimization Through Stochastic Sampling

Abstract

We present a new algorithm to optimize distributions defined implicitly by parameterized stochastic diffusions. Doing so allows us to modify the outcome distribution of sampling processes by optimizing over their parameters. We introduce a general framework for first-order optimization of these processes, that performs jointly, in a single loop, optimization and sampling steps. We showcase it in training and finetuning applications.

Cite

Text

Marion et al. "Implicit Diffusion: Efficient Optimization Through Stochastic Sampling." ICML 2024 Workshops: Differentiable_Almost_Everything, 2024.

Markdown

[Marion et al. "Implicit Diffusion: Efficient Optimization Through Stochastic Sampling." ICML 2024 Workshops: Differentiable_Almost_Everything, 2024.](https://mlanthology.org/icmlw/2024/marion2024icmlw-implicit/)

BibTeX

@inproceedings{marion2024icmlw-implicit,
  title     = {{Implicit Diffusion: Efficient Optimization Through Stochastic Sampling}},
  author    = {Marion, Pierre and Korba, Anna and Bartlett, Peter and Blondel, Mathieu and De Bortoli, Valentin and Doucet, Arnaud and Llinares-López, Felipe and Paquette, Courtney and Berthet, Quentin},
  booktitle = {ICML 2024 Workshops: Differentiable_Almost_Everything},
  year      = {2024},
  url       = {https://mlanthology.org/icmlw/2024/marion2024icmlw-implicit/}
}