Pareto Policy Pool for Model-based Offline Reinforcement Learning.
Yijun YangJing JiangTianyi ZhouJie MaYuhui ShiPublished in: ICLR (2022)
Keyphrases
- reinforcement learning
- optimal policy
- model free
- policy search
- policy iteration
- markov decision process
- action selection
- markov decision processes
- state space
- multi objective
- policy evaluation
- function approximators
- reinforcement learning algorithms
- partially observable environments
- function approximation
- policy gradient
- reward function
- markov decision problems
- control policies
- real time
- partially observable
- control policy
- approximate dynamic programming
- reinforcement learning problems
- rl algorithms
- state and action spaces
- machine learning
- actor critic
- average reward
- action space
- partially observable markov decision processes
- genetic algorithm
- multi objective optimization
- state action
- multiobjective optimization
- optimal control
- agent receives
- infinite horizon
- control problems
- pareto optimal
- state dependent
- transition model
- temporal difference
- sufficient conditions
- supervised learning
- multicriteria optimization
- average cost
- partially observable domains
- policy gradient methods