Your Diffusion Model is Secretly a Noise Classifier and Benefits from Contrastive Training
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915002306265088 |
|---|---|
| author | Wu, Yunshu Luo, Yingtao Kong, Xianghao Papalexakis, Evangelos E. Steeg, Greg Ver |
| author_facet | Wu, Yunshu Luo, Yingtao Kong, Xianghao Papalexakis, Evangelos E. Steeg, Greg Ver |
| contents | Diffusion models learn to denoise data and the trained denoiser is then used to generate new samples from the data distribution. In this paper, we revisit the diffusion sampling process and identify a fundamental cause of sample quality degradation: the denoiser is poorly estimated in regions that are far Outside Of the training Distribution (OOD), and the sampling process inevitably evaluates in these OOD regions. This can become problematic for all sampling methods, especially when we move to parallel sampling which requires us to initialize and update the entire sample trajectory of dynamics in parallel, leading to many OOD evaluations. To address this problem, we introduce a new self-supervised training objective that differentiates the levels of noise added to a sample, leading to improved OOD denoising performance. The approach is based on our observation that diffusion models implicitly define a log-likelihood ratio that distinguishes distributions with different amounts of noise, and this expression depends on denoiser performance outside the standard training distribution. We show by diverse experiments that the proposed contrastive diffusion training is effective for both sequential and parallel settings, and it improves the performance and speed of parallel samplers significantly. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_08946 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Your Diffusion Model is Secretly a Noise Classifier and Benefits from Contrastive Training Wu, Yunshu Luo, Yingtao Kong, Xianghao Papalexakis, Evangelos E. Steeg, Greg Ver Machine Learning Diffusion models learn to denoise data and the trained denoiser is then used to generate new samples from the data distribution. In this paper, we revisit the diffusion sampling process and identify a fundamental cause of sample quality degradation: the denoiser is poorly estimated in regions that are far Outside Of the training Distribution (OOD), and the sampling process inevitably evaluates in these OOD regions. This can become problematic for all sampling methods, especially when we move to parallel sampling which requires us to initialize and update the entire sample trajectory of dynamics in parallel, leading to many OOD evaluations. To address this problem, we introduce a new self-supervised training objective that differentiates the levels of noise added to a sample, leading to improved OOD denoising performance. The approach is based on our observation that diffusion models implicitly define a log-likelihood ratio that distinguishes distributions with different amounts of noise, and this expression depends on denoiser performance outside the standard training distribution. We show by diverse experiments that the proposed contrastive diffusion training is effective for both sequential and parallel settings, and it improves the performance and speed of parallel samplers significantly. |
| title | Your Diffusion Model is Secretly a Noise Classifier and Benefits from Contrastive Training |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2407.08946 |