arXiv：cs.LG· Hanru Bai, Yuanchao Xu, Fengyi Li·· 4 小时前AI 评分33
Koopman 观测器加速扩散模型：用浅层特征校正深层特征预测
Koopman Observers for Diffusion Acceleration: Correcting Feature Forecasts with Shallow Measurements
AI 导读
研究者提出一种观测校正的 Koopman 框架，可在不改变冻结扩散模型参数的前提下加速采样：用浅层网络特征的实时测量来校正深层特征的预测。在 CIFAR-10 和 ImageNet 十类子集上（各 3 次 10,000 张图像运行），配对 Inception 特征 MSE 相对通道仿射预测分别降低 19.9% 和 11.9%，其中观测校正贡献 4.54% 和 4.67%。
正文
Abstract:Feature caching accelerates diffusion sampling by replacing expensive network evaluations with predictions from previously computed activations. However, forecasts based only on past features cannot directly incorporate changes in the current denoising state. We investigate whether inexpensive, freshly computed features can serve as observations for correcting these predictions. We introduce an observation-corrected Koopman framework for accelerating frozen diffusion models. Using calibration trajectories, we identify finite-dimensional, time-dependent Koopman approximations that jointly describe the increments of shallow and deep network features. During accelerated sampling, these operators predict the evolution of expensive deep features, while innovations in the observed shallow features correct the predicted state. Periodic full evaluations refresh the observer, and all generative-model parameters remain unchanged. This formulation enables controlled comparisons of temporal prediction and observation correction. Across three 10,000-image runs per dataset, our method reduces paired Inception-feature MSE by $19.9\%$ on CIFAR-10 and $11.9\%$ on a ten-class ImageNet subset relative to channelwise affine prediction under the same four-partial-step schedule. Matched ablations attribute additional reductions of $4.54\%$ and $4.67\%$ to observation correction. The observer achieves $1.89\times$ and $1.85\times$ measured speedups over DDIM-50, supporting improved reference-sampler fidelity without retraining the denoiser.
| Comments: | 14 pages |
| Subjects: | Machine Learning (cs.LG); Dynamical Systems (math.DS); Machine Learning (stat.ML) |
| Cite as: | arXiv:2610.10366 [cs.LG] |
| (or arXiv:2610.10366v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.10366 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Fengyi Li [view email]
[v1]
Wed, 7 Oct 2026 16:35:07 UTC (418 KB)
来源：arXiv：cs.LG · arxiv.org