Reproduce “On the Identification and Mitigation of Weaknesses in the Knowledge Gradient Policy for Multi-Armed 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
James Edwards, Paul Fearnhead, Kevin Glazebrook
The authors analyze Knowledge Gradient (KG), which scores arms by current reward and the expected change in the best estimated mean after one observation. They prove that with rewards bounded below, KG can choose an arm worse in both mean and information than another; Gaussian rewards avoid this error. Proposed repairs prevent such choices, but do not guarantee optimal reward.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.