Policy-regularized Offline Multi-objective Reinforcement Learning.
Qian LinChao YuZongkai LiuZifan WuPublished in: CoRR (2024)
Keyphrases
- multi objective
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- evolutionary algorithm
- action selection
- objective function
- multi objective optimization
- optimization algorithm
- reinforcement learning algorithms
- partially observable environments
- state space
- genetic algorithm
- reinforcement learning problems
- partially observable
- function approximation
- control policies
- markov decision processes
- state and action spaces
- markov decision problems
- particle swarm optimization
- multiple objectives
- approximate dynamic programming
- action space
- actor critic
- control policy
- function approximators
- policy iteration
- state action
- reward function
- least squares
- policy gradient
- policy evaluation
- average reward
- rl algorithms
- real time
- temporal difference
- multi objective optimization problems
- decision problems
- dynamic programming
- partially observable domains
- continuous state spaces
- conflicting objectives
- nsga ii
- model free
- pareto optimal
- risk minimization
- multi agent
- partially observable markov decision process
- exploration exploitation tradeoff
- state dependent
- infinite horizon
- optimal control
- learning algorithm
- machine learning
- continuous state
- partially observable markov decision processes
- average cost
- long run
- image restoration
- optimization problems
- agent receives