Tree-based Fitted Q-iteration for Multi-Objective Markov Decision problems.
Andrea CastellettiFrancesca PianosiMarcello RestelliPublished in: IJCNN (2012)
Keyphrases
- markov decision problems
- multi objective
- fitted q iteration
- reinforcement learning
- evolutionary algorithm
- partially observable
- state space
- multi objective optimization
- optimization algorithm
- pareto optimal
- expected utility
- linear programming
- optimal policy
- objective function
- function approximation
- multiple objectives
- policy iteration
- decision theoretic
- dynamic programming
- markov decision processes
- genetic algorithm
- nsga ii
- machine learning
- model free
- learning algorithm
- supervised learning
- action space
- decision problems
- utility function
- temporal difference
- reward function
- hidden markov models
- optimal solution
- average cost
- multi agent
- function approximators
- decision processes
- neural network