Reproduce “Meta-learning for mixed linear regression” 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
Weihao Kong, Raghav Somani, Zhao Song, Sham Kakade, Sewoong Oh
The authors ask whether many small tasks can replace few labeled tasks for finite mixtures of k linear regressions. Their spectral method uses light tasks to estimate shared structure, medium tasks to cluster components, and a second light batch to refine estimates. Under stated assumptions, roughly dk² light-task examples and k^(3/2) medium tasks with √k examples each suffice; this is a sufficient regime, not a universal bound.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.