Reproduce “Enabling risk-aware Reinforcement Learning for medical interventions through uncertainty decomposition” 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
Paul Festor, Giulia Luise, Matthieu Komorowski, A. Aldo Faisal
The authors adapt distributional quantile-regression DQN with Bayesian inference to separate uncertainty in reinforcement learning. The method estimates epistemic uncertainty from variation across model parameters and aleatoric uncertainty from variation across return quantiles. In grid worlds, estimates respond to sparse data and environmental randomness. In a MIMIC-III-derived sepsis MDP, epistemic uncertainty is inversely associated with state visitation; this retrospective proof of concept does not establish clinical benefit.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.