Reproduce “A Batch, Off-Policy, Actor-Critic Algorithm for Optimizing the Average Reward” 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
S.A. Murphy, Y. Deng, E.B. Laber, H.R. Maei, R.S. Sutton, K. Witkiewitz
Murphy et al. propose batch off-policy actor–critic learning of an interpretable, stochastic rule for when to offer mobile-health treatment, using fixed histories from multiple people. In simulation, the learned policy’s reward tracked the model-computed optimum, while always treating lost about 50% as burden increased from τ = 0.20 to 0.60. BASICS-Mobile fits a rule, not evidence of improved smoking outcomes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.