Reproduce “Thompson Sampling with Unrestricted Delays” 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
Han Wu, Stefan Wager
The authors analyze ordinary Thompson Sampling for Bernoulli bandits with delayed rewards, sampling each arm’s reward rate from its current posterior and updating beliefs when feedback arrives. For two arms, they prove a regret bound based on selected delay quantiles, matching the prior method’s order. Simulations show lower regret than delay-aware baselines under i.i.d. delays independent of rewards.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.