Exploiting Action-Value Uncertainty to Drive Exploration in Reinforcement Learning.
Carlo D'EramoAndrea CiniMarcello RestelliPublished in: IJCNN (2019)
Keyphrases
- action selection
- reinforcement learning
- active exploration
- exploration strategy
- action space
- reward shaping
- model based reinforcement learning
- partially observable domains
- function approximation
- markov decision processes
- autonomous learning
- state action
- temporal difference
- model free
- exploration exploitation
- optimal policy
- partial observability
- uncertain data
- learning process
- machine learning
- learning problems
- sensing actions
- state space
- transition model
- exploration exploitation tradeoff
- reinforcement learning algorithms
- decision theory
- human actions
- learning algorithm
- continuous state
- inherent uncertainty
- expected utility
- optimal control
- conditional probabilities
- dynamic programming