Automatic basis function construction for approximate dynamic programming and reinforcement learning.
Philipp W. KellerShie MannorDoina PrecupPublished in: ICML (2006)
Keyphrases
- approximate dynamic programming
- factored mdps
- reinforcement learning
- basis functions
- state space
- linear program
- stochastic dynamic programming
- dynamic programming
- linear combination
- control policy
- markov decision processes
- policy iteration
- actor critic
- linear programming
- radial basis function
- function approximation
- temporal difference
- continuous state
- step size
- reinforcement learning algorithms
- multiresolution
- optimal policy
- machine learning
- special case
- evolutionary algorithm
- multiscale
- learning algorithm