Multi-objective reinforcement learning using sets of pareto dominating policies.
Kristof Van MoffaertAnn NowéPublished in: J. Mach. Learn. Res. (2014)
Keyphrases
- multi objective
- multiobjective optimization
- reinforcement learning
- optimal policy
- multi objective optimization
- evolutionary algorithm
- optimization algorithm
- policy search
- genetic algorithm
- markov decision process
- multiple objectives
- objective function
- pareto optimal
- conflicting objectives
- particle swarm optimization
- nsga ii
- reward function
- control policies
- multi agent
- markov decision processes
- pareto optimal solutions
- multi objective optimization problems
- bi objective
- model free
- finite state
- learning algorithm
- state space
- multi objective evolutionary
- reinforcement learning agents
- hierarchical reinforcement learning
- partially observable markov decision processes
- optimum design
- differential evolution
- continuous state
- fitted q iteration
- temporal difference
- learning process
- dynamic programming
- policy iteration
- function approximation
- machine learning
- average cost
- reinforcement learning algorithms
- action selection