Reproduce “Making Look-Ahead Active Learning Strategies Feasible with Neural Tangent Kernels” 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
Mohamad Amin Mohamadi, Wonho Bae, Danica J. Sutherland
The authors use an empirical neural tangent kernel (NTK) to estimate how labeling a candidate changes predictions across an unlabeled pool, avoiding retraining for every candidate. On MNIST, the method takes over 100 times less wall-clock time than a 15-epoch naïve look-ahead baseline and achieves higher accuracy in the reported experiment; it remains slower than common myopic methods.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.