Reproduce “Subsampling for supervised learning in reproducing kernel Hilbert spaces” 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
Eyal Vayness, Maxime Sangnier
The authors develop a pilot-based subsampling method for kernel learning, selecting observations by loss sensitivity and kernel diagonal, then correcting unequal selection with probability weights. Theory gives an asymptotically optimal rule. Experiments find greater accuracy than uniform subsampling at roughly similar training times, and greater efficiency than kernel approximations on Covertype, but not on two synthetic tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.