SRC-2105.05233
Source
扩散超越 GAN 的系统改进(Dhariwal & Nichol, NeurIPS 2021);改进采样器与架构。
| id | SRC-2105.05233 |
|---|---|
| type | source |
| kind | paper |
| title | Diffusion Models Beat GANs on Image Synthesis |
| authors | Prafulla Dhariwal, Alex Nichol |
| venue | NeurIPS 2021 |
| arxiv | 2105.05233 |
| tier | 0 |
| lifecycle | EXTRACTED |
| epistemic | n/a |
| ingested | 2026-09-03 |
| version | arXiv:2105.05233 |
| source-hash | sha256:0000000000000000000000000000000000000000000000000000000000000000 |
| source-role | g2dp-citation-neighbor |
| g2dp-anchor | SRC-2606.26017 |
| g2dp-relation | CITED_BY_G2DP | G2DP 引用本工作作为扩散模型改进与高效采样(Dhariwal & Nichol, NeurIPS 2021)的脉络;与 DPM-Solver 共同构成低步数采样技术链。 |
| citation-count-s2 | 12850 |
对扩散采样与生成质量的系统改进(Dhariwal & Nichol, NeurIPS 2021):引入改进采样器与架构调整,使扩散在图像合成上超越 GAN。
与 G2DP 的关系(SRC-2606.26017):G2DP 引用本工作作为扩散模型改进与高效采样(Dhariwal & Nichol, NeurIPS 2021)的脉络;与 DPM-Solver 共同构成低步数采样技术链。
引用(1)
- SRC-2606.26017 G2DP:时空占据栅格+路线进度融合为可微 cost volume,去噪末段注入引导,nuPlan 纯模仿 SOTA;本引用簇中心。