Adaptive Destruction Processes for Diffusion Samplers
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912409463029760 |
|---|---|
| author | Gritsaev, Timofei Morozov, Nikita Tamogashev, Kirill Tiapkin, Daniil Samsonov, Sergey Naumov, Alexey Vetrov, Dmitry Malkin, Nikolay |
| author_facet | Gritsaev, Timofei Morozov, Nikita Tamogashev, Kirill Tiapkin, Daniil Samsonov, Sergey Naumov, Alexey Vetrov, Dmitry Malkin, Nikolay |
| contents | This paper explores the challenges and benefits of a trainable destruction process in diffusion samplers -- diffusion-based generative models trained to sample an unnormalised density without access to data samples. Contrary to the majority of work that views diffusion samplers as approximations to an underlying continuous-time model, we view diffusion models as discrete-time policies trained to produce samples in very few generation steps. We propose to trade some of the elegance of the underlying theory for flexibility in the definition of the generative and destruction policies. In particular, we decouple the generation and destruction variances, enabling both transition kernels to be learned as unconstrained Gaussian densities. We show that, when the number of steps is limited, training both generation and destruction processes results in faster convergence and improved sampling quality on various benchmarks. Through a robust ablation study, we investigate the design choices necessary to facilitate stable training. Finally, we show the scalability of our approach through experiments on GAN latent space sampling for conditional image generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_01541 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Adaptive Destruction Processes for Diffusion Samplers Gritsaev, Timofei Morozov, Nikita Tamogashev, Kirill Tiapkin, Daniil Samsonov, Sergey Naumov, Alexey Vetrov, Dmitry Malkin, Nikolay Machine Learning This paper explores the challenges and benefits of a trainable destruction process in diffusion samplers -- diffusion-based generative models trained to sample an unnormalised density without access to data samples. Contrary to the majority of work that views diffusion samplers as approximations to an underlying continuous-time model, we view diffusion models as discrete-time policies trained to produce samples in very few generation steps. We propose to trade some of the elegance of the underlying theory for flexibility in the definition of the generative and destruction policies. In particular, we decouple the generation and destruction variances, enabling both transition kernels to be learned as unconstrained Gaussian densities. We show that, when the number of steps is limited, training both generation and destruction processes results in faster convergence and improved sampling quality on various benchmarks. Through a robust ablation study, we investigate the design choices necessary to facilitate stable training. Finally, we show the scalability of our approach through experiments on GAN latent space sampling for conditional image generation. |
| title | Adaptive Destruction Processes for Diffusion Samplers |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2506.01541 |