Reproduce “Meta-learning with Stochastic Linear Bandits” 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

Leonardo Cella, Alessandro Lazaric, Massimiliano Pontil

The authors adapt meta-learning’s shared-bias idea to stochastic linear bandits, regularizing a new task’s reward estimate toward a bias learned from earlier tasks rather than zero. They propose averaging ridge estimates or pooling samples to learn it. In Last.fm tests on 20 selected similar users, both methods had lower cumulative regret than independent-task learning, a result limited to that setting.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.