Interpretable Multi-Objective Reinforcement Learning through Policy Orchestration.
Ritesh NoothigattuDjallel BouneffoufNicholas MatteiRachita ChandraPiyush MadanKush R. VarshneyMurray CampbellMoninder SinghFrancesca RossiPublished in: CoRR (2018)
Keyphrases
- multi objective
- reinforcement learning
- optimal policy
- policy search
- evolutionary algorithm
- markov decision process
- action selection
- multi objective optimization
- optimization algorithm
- function approximation
- state space
- function approximators
- partially observable environments
- action space
- particle swarm optimization
- markov decision problems
- multiobjective optimization
- approximate dynamic programming
- markov decision processes
- reinforcement learning problems
- actor critic
- policy iteration
- policy evaluation
- policy gradient
- temporal difference
- reward function
- objective function
- genetic algorithm
- control policy
- reinforcement learning algorithms
- control policies
- state and action spaces
- web services
- multiple objectives
- nsga ii
- state action
- decision problems
- machine learning
- multi objective optimization problems
- learning algorithm
- dynamic programming
- partially observable
- finite state
- infinite horizon
- state dependent
- optimum design
- continuous state spaces
- pareto optimal
- average reward
- partially observable markov decision processes
- model free
- optimal control
- multi agent
- agent learns
- continuous state
- long run
- rl algorithms
- bi objective
- multi objective evolutionary algorithms
- transition model
- search space
- neural network
- policy gradient methods
- multiagent evolutionary algorithm