Biased Aggregation, Rollout, and Enhanced Policy Improvement for Reinforcement Learning.
Dimitri P. BertsekasPublished in: CoRR (2019)
Keyphrases
- reinforcement learning
- approximate policy iteration
- optimal policy
- policy search
- markov decision process
- policy iteration
- action selection
- reinforcement learning algorithms
- markov decision processes
- state space
- markov decision problems
- temporal difference
- reinforcement learning problems
- control policy
- actor critic
- function approximators
- policy gradient
- control policies
- function approximation
- markov games
- partially observable environments
- machine learning
- average reward
- reward function
- state and action spaces
- continuous state
- policy evaluation
- neural network
- policy gradient methods
- rl algorithms
- model free
- multi agent
- temporal difference learning
- action space
- data aggregation
- control problems
- average cost
- partially observable
- finite state
- optimal control
- learning problems
- decision problems
- significant improvement
- partially observable domains
- high quality