Reproduce “Koopman Observers for Diffusion Acceleration: Correcting Feature Forecasts with Shallow Measurements” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hanru Bai, Yuanchao Xu, Fengyi Li
The method accelerates diffusion sampling by using shallow features to correct forecasts of costly deep features, while keeping denoiser weights frozen. Offline-fitted predictors and correction gains support partial evaluations between full refreshes, without changing the 50-step grid. At four partial steps, paired Inception-feature MSE falls 19.9% on CIFAR-10 and 11.9% on ImageNet-10, with speedups of 1.89× and 1.85× over DDIM-50.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.