Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2402.13144 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913630442749952 |
|---|---|
| author | Wang, Kai Tang, Dongwen Zeng, Boya Yin, Yida Xu, Zhaopan Zhou, Yukun Zang, Zelin Darrell, Trevor Liu, Zhuang You, Yang |
| author_facet | Wang, Kai Tang, Dongwen Zeng, Boya Yin, Yida Xu, Zhaopan Zhou, Yukun Zang, Zelin Darrell, Trevor Liu, Zhuang You, Yang |
| contents | Diffusion models have achieved remarkable success in image and video generation. In this work, we demonstrate that diffusion models can also \textit{generate high-performing neural network parameters}. Our approach is simple, utilizing an autoencoder and a diffusion model. The autoencoder extracts latent representations of a subset of the trained neural network parameters. Next, a diffusion model is trained to synthesize these latent representations from random noise. This model then generates new representations, which are passed through the autoencoder's decoder to produce new subsets of high-performing network parameters. Across various architectures and datasets, our approach consistently generates models with comparable or improved performance over trained networks, with minimal additional cost. Notably, we empirically find that the generated models are not memorizing the trained ones. Our results encourage more exploration into the versatile use of diffusion models. Our code is available \href{https://github.com/NUS-HPC-AI-Lab/Neural-Network-Diffusion}{here}. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_13144 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Neural Network Diffusion Wang, Kai Tang, Dongwen Zeng, Boya Yin, Yida Xu, Zhaopan Zhou, Yukun Zang, Zelin Darrell, Trevor Liu, Zhuang You, Yang Machine Learning Computer Vision and Pattern Recognition Diffusion models have achieved remarkable success in image and video generation. In this work, we demonstrate that diffusion models can also \textit{generate high-performing neural network parameters}. Our approach is simple, utilizing an autoencoder and a diffusion model. The autoencoder extracts latent representations of a subset of the trained neural network parameters. Next, a diffusion model is trained to synthesize these latent representations from random noise. This model then generates new representations, which are passed through the autoencoder's decoder to produce new subsets of high-performing network parameters. Across various architectures and datasets, our approach consistently generates models with comparable or improved performance over trained networks, with minimal additional cost. Notably, we empirically find that the generated models are not memorizing the trained ones. Our results encourage more exploration into the versatile use of diffusion models. Our code is available \href{https://github.com/NUS-HPC-AI-Lab/Neural-Network-Diffusion}{here}. |
| title | Neural Network Diffusion |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2402.13144 |