Ensemble Usage for More Reliable Policy Identification in Reinforcement Learning.
Alexander HansSteffen UdluftPublished in: ESANN (2011)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- markov decision process
- learning algorithm
- partially observable environments
- reinforcement learning algorithms
- state space
- ensemble methods
- reward function
- markov decision processes
- multi agent
- control policies
- approximate dynamic programming
- policy evaluation
- control policy
- inverse reinforcement learning
- partially observable domains
- ensemble learning
- reinforcement learning problems
- random forests
- partially observable
- actor critic
- rl algorithms
- cost effective
- action space
- learning process
- policy gradient
- dynamic programming
- state action
- function approximators
- policy iteration
- neural network
- optimal control
- feature selection
- data sets
- state and action spaces
- infinite horizon
- training data
- continuous state spaces
- training set
- function approximation
- control problems
- model free