Energy-Based Policy Constraint for Offline Reinforcement Learning.
Zhiyong PengChanglin HanYadong LiuZongtan ZhouPublished in: CICAI (2) (2023)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- markov decision process
- partially observable environments
- reinforcement learning problems
- partially observable
- markov decision processes
- state space
- function approximation
- state and action spaces
- function approximators
- policy gradient
- control policy
- rl algorithms
- reinforcement learning algorithms
- reward function
- global constraints
- approximate dynamic programming
- control policies
- action space
- state action
- policy evaluation
- actor critic
- markov decision problems
- continuous state spaces
- robotic control
- real time
- state dependent
- optimal control
- decision problems
- policy iteration
- machine learning
- dynamic programming
- model free
- finite state
- infinite horizon
- sufficient conditions
- learning problems
- agent learns
- neural network
- learning classifier systems
- partially observable markov decision processes
- reinforcement learning methods
- temporal difference learning