A performance gradient perspective on gradient-based policy iteration and a modified value iteration.
Lei YangJames DankertJennie SiPublished in: Int. J. Intell. Comput. Cybern. (2008)
Keyphrases
- policy iteration
- model free
- markov decision processes
- reinforcement learning
- fixed point
- sample path
- optimal policy
- least squares
- approximate dynamic programming
- markov decision process
- finite state
- temporal difference
- average reward
- policy evaluation
- reinforcement learning algorithms
- infinite horizon
- dynamic programming
- linear programming
- function approximation
- state space
- neural network
- partially observable
- optimal control
- policy gradient
- average cost
- optimal solution
- markov decision problems
- evaluation function