The divergence of reinforcement learning algorithms with value-iteration and function approximation.
Michael FairbankEduardo AlonsoPublished in: IJCNN (2012)
Keyphrases
- reinforcement learning algorithms
- function approximation
- reinforcement learning
- markov decision processes
- model free
- policy iteration
- state space
- temporal difference
- reinforcement learning problems
- optimal policy
- average reward
- markov decision process
- partially observable markov decision processes
- reinforcement learning methods
- dynamic programming
- temporal difference learning
- function approximators
- radial basis function
- finite state
- policy search
- td learning
- infinite horizon
- learning tasks
- reward function
- policy gradient
- policy evaluation
- action selection
- learning agent
- control problems
- machine learning
- state variables
- optimal control
- basis functions
- text classification
- supervised learning
- decision trees