Addressing Value Estimation Errors in Reinforcement Learning with a State-Action Return Distribution Function.
Jingliang DuanYang GuanYangang RenShengbo Eben LiBo ChengPublished in: CoRR (2020)
Keyphrases
- state action
- distribution function
- estimation error
- reinforcement learning
- evaluation function
- random variables
- markov decision process
- action space
- function approximators
- state space
- error rate
- standard deviation
- function approximation
- markov decision processes
- covariance matrix
- stochastic games
- learning algorithm
- average reward
- machine learning
- optimal policy
- reinforcement learning algorithms
- dynamic programming
- state transitions
- optimal control
- action selection
- temporal difference
- multiresolution
- density function
- supervised learning
- multi agent