Expected Policy Gradients for Reinforcement Learning.
Kamil CiosekShimon WhitesonPublished in: J. Mach. Learn. Res. (2020)
Keyphrases
- reinforcement learning
- optimal policy
- total reward
- policy search
- action selection
- partially observable environments
- markov decision process
- markov decision processes
- function approximators
- reinforcement learning algorithms
- function approximation
- control policies
- partially observable
- policy gradient
- control policy
- markov decision problems
- approximate dynamic programming
- state and action spaces
- state space
- decision problems
- actor critic
- action space
- average reward
- reward function
- reinforcement learning problems
- dynamic programming
- policy iteration
- policy evaluation
- opportunity cost
- policy gradient methods
- continuous state spaces
- transition model
- learning algorithm
- continuous state
- state action
- asymptotically optimal
- control problems
- infinite horizon
- optimal control
- learning process
- neural network
- stationary policies
- state dependent
- image gradient
- temporal difference
- model free
- markov chain
- model free reinforcement learning
- exploration exploitation tradeoff