Reproduce “A Theory of Adversary-Directed Online Learning” 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
Steve Hanneke, Amirreza Shaeiri
The authors study learning where the learner knows a pool of instances but an adversary chooses their order. In realizable learning, mistakes are constant with finite Littlestone dimension, logarithmic when it is infinite but perfect Littlestone dimension is finite, and linear when perfect Littlestone dimension is infinite. For agnostic regret, rates are tilde-Theta square root T with finite perfect Littlestone dimension and linear with infinite dimension. These are theoretical results.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.