用约10个参数优化扩散模型采样，速度更快质量更高
Diffusion Sampling Correction via Approximately 10 Parameters
- 通过PCA提取采样空间主成分，仅需学习少量坐标修正方向
- 在CIFAR10上将DDIM的FID从15.69降至4.37（10次推理）
- 仅需12个参数、单卡几分钟训练，可直接接入现有模型
尽管生成能力强大，扩散概率模型（DPMs）面临采样速度慢的问题。现有基于知识蒸馏的方法通常需要大量额外训练成本和参数存储，限制了实用性。本文提出基于PCA的自适应搜索（PAS），以极低开销优化现有求解器。首先利用PCA获取少数基向量以覆盖高维采样空间，从而只需学习一组坐标来修正采样方向；进一步观察到累积截断误差呈' S'形，设计自适应搜索策略，显著提升采样效率并将存储参数数降至约10个。大量实验表明，PAS可无痛增强现有快速求解器，代价极小。例如在CIFAR10上，使用仅12个参数、单张A100 GPU不到一分钟训练，将DDIM的FID从15.69降至4.37（NFE=10）。代码已开源。
原文摘要 · Abstract (English)
While powerful for generation, Diffusion Probabilistic Models (DPMs) face slow sampling challenges, for which various distillation-based methods have been proposed. However, they typically require significant additional training costs and model parameter storage, limiting their practicality. In this work, we propose PCA-based Adaptive Search (PAS), which optimizes existing solvers for DPMs with minimal additional costs. Specifically, we first employ PCA to obtain a few basis vectors to span the high-dimensional sampling space, which enables us to learn just a set of coordinates to correct the sampling direction; furthermore, based on the observation that the cumulative truncation error exhibits an ``S"-shape, we design an adaptive search strategy that further enhances the sampling efficiency and reduces the number of stored parameters to approximately 10. Extensive experiments demonstrate that PAS can significantly enhance existing fast solvers in a plug-and-play manner with negligible costs. E.g., on CIFAR10, PAS optimizes DDIM's FID from 15.69 to 4.37 (NFE=10) using only 12 parameters and sub-minute training on a single A100 GPU. Code is available at https://github.com/onefly123/PAS.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。