Reward-Free Policy Space Compression for Reinforcement Learning.
Mirco MuttiStefano Del ColMarcello RestelliPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- optimal policy
- action space
- reward function
- partially observable environments
- average reward
- markov decision processes
- state space
- policy search
- inverse reinforcement learning
- policy gradient
- action selection
- total reward
- reinforcement learning algorithms
- eligibility traces
- state action
- markov decision process
- function approximators
- machine learning
- approximate dynamic programming
- policy iteration
- function approximation
- model free
- reinforcement learning methods
- data compression
- reinforcement learning problems
- control policy
- control policies
- partially observable
- space time
- actor critic
- optimal control
- multi agent
- agent learns
- state and action spaces
- video sequences
- low dimensional
- learning process
- continuous state spaces
- markov decision problems
- rl algorithms
- search space
- state dependent
- learning algorithm