Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2411.09356 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916480023527424 |
|---|---|
| author | Xiao, Xiongye Li, Shixuan Huang, Luzhe Liu, Gengshuo Nguyen, Trung-Kien Huang, Yi Chang, Di Kochenderfer, Mykel J. Bogdan, Paul |
| author_facet | Xiao, Xiongye Li, Shixuan Huang, Luzhe Liu, Gengshuo Nguyen, Trung-Kien Huang, Yi Chang, Di Kochenderfer, Mykel J. Bogdan, Paul |
| contents | While working within the spatial domain can pose problems associated with ill-conditioned scores caused by power-law decay, recent advances in diffusion-based generative models have shown that transitioning to the wavelet domain offers a promising alternative. However, within the wavelet domain, we encounter unique challenges, especially the sparse representation of high-frequency coefficients, which deviates significantly from the Gaussian assumptions in the diffusion process. To this end, we propose a multi-scale generative modeling in the wavelet domain that employs distinct strategies for handling low and high-frequency bands. In the wavelet domain, we apply score-based generative modeling with well-conditioned scores for low-frequency bands, while utilizing a multi-scale generative adversarial learning for high-frequency bands. As supported by the theoretical analysis and experimental results, our model significantly improve performance and reduce the number of trainable parameters, sampling steps, and time. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2411_09356 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Multi-scale Generative Modeling for Fast Sampling Xiao, Xiongye Li, Shixuan Huang, Luzhe Liu, Gengshuo Nguyen, Trung-Kien Huang, Yi Chang, Di Kochenderfer, Mykel J. Bogdan, Paul Artificial Intelligence While working within the spatial domain can pose problems associated with ill-conditioned scores caused by power-law decay, recent advances in diffusion-based generative models have shown that transitioning to the wavelet domain offers a promising alternative. However, within the wavelet domain, we encounter unique challenges, especially the sparse representation of high-frequency coefficients, which deviates significantly from the Gaussian assumptions in the diffusion process. To this end, we propose a multi-scale generative modeling in the wavelet domain that employs distinct strategies for handling low and high-frequency bands. In the wavelet domain, we apply score-based generative modeling with well-conditioned scores for low-frequency bands, while utilizing a multi-scale generative adversarial learning for high-frequency bands. As supported by the theoretical analysis and experimental results, our model significantly improve performance and reduce the number of trainable parameters, sampling steps, and time. |
| title | Multi-scale Generative Modeling for Fast Sampling |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2411.09356 |