Enabling risk-aware Reinforcement Learning for medical interventions through uncertainty decomposition.
Paul FestorGiulia LuiseMatthieu KomorowskiA. Aldo FaisalPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- risk measures
- risk averse
- expected utility
- partial observability
- medical domain
- function approximation
- medical imaging
- medical information
- state space
- optimal policy
- reinforcement learning algorithms
- high risk
- medical data
- learning problems
- markov decision processes
- utility function
- decomposition method
- portfolio management
- risk aversion
- probability theory
- machine learning
- uncertain data
- dynamic programming
- extreme events
- model free
- decomposition algorithm
- medical experts
- risk assessment
- medical diagnosis
- supervised learning
- decision making