| Article ID: | iaor20011955 |
| Country: | Hungary |
| Volume: | 19 |
| Issue: | 1 |
| Start Page Number: | 39 |
| End Page Number: | 68 |
| Publication Date: | Jan 1999 |
| Journal: | Alkalmazott Matematikai Lapok |
| Authors: | Szepesvri Csaba |
| Keywords: | markov processes |
In the paper we are concerned with the adaptive optimal control of Markov decision problems. An asynchronous stochastic approximation theorem is proved. Then the theory of Markov decision problems is reviewed and three applications of the main theorem are given for the estimation of the optimal cost-function. Finally, a class of adaptive optimal policies is constructed.