A Necessary Condition of Convergence for Reinforcement Learning with Function Approximation.
Artur MerkeRalf SchoknechtPublished in: ICML (2002)
Keyphrases
- function approximation
- reinforcement learning
- temporal difference
- tile coding
- temporal difference learning
- state action space
- mountain car
- model free
- function approximators
- temporal difference learning algorithms
- reinforcement learning algorithms
- radial basis function
- learning tasks
- sufficient conditions
- td learning
- convergence rate
- learning algorithm
- convergence speed
- initial conditions
- genetic algorithm
- machine learning
- transfer learning
- text classification
- markov decision processes
- supervised learning
- state space
- artificial neural networks
- pattern recognition
- multi agent