Reproduce “Random Projections for Spectral Algorithms: Optimal Rates in Sobolev Norms” 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
Caixing Wang, Yue Wang
The authors analyze random projection for kernel regression, which reduces computation by restricting estimators to lower-dimensional subspaces. They prove minimax-optimal Sobolev-norm rates up to logarithmic factors under stated assumptions, when target smoothness exceeds a threshold and projection error is sufficiently small. In a synthetic rough-target example with s=0.5, projected methods’ learning curves track predicted rates and resemble full methods.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.