Reproduce “Score Approximation for Diffusion Models on Arbitrary Low-Dimensional Structures” 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
Xinhe Mu, Zaijiu Shang, Zhaoqi Zhou, Chuan Zhou, Qi Meng, Guiying Yan, Zhiming Ma
The authors show that a ReLU network can approximate the score of Gaussian-noised compactly supported data without assuming a smooth density. They cover the support with local pieces, approximate each piece, and combine estimates by their fit to the noisy input. Under variance-exploding noise, squared error is at most 6ε²; synthetic tests show similar width trends for distributions sharing support dimension.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.