Reproduce “Convergence Rates of Active Learning for Maximum Likelihood Estimation” 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
Kamalika Chaudhuri, Sham Kakade, Praneeth Netrapalli, Sujay Sanghavi
The authors develop an active learning method for maximum likelihood estimation that selects labels to reduce expected excess negative log-likelihood on a fixed unlabeled pool. It first uses random labels to estimate curvature, then chooses a targeted batch. Under stated assumptions, theory gives a matching leading rate; one constructed regression pool has rates d/m versus at most 4/m.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.