Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Classifier-based policy representation

Rexakis Ioannis, Lagoudakis Michael

Full record


URI: http://purl.tuc.gr/dl/dias/2E55B7D4-6FCA-4907-8055-F24FEEF56CC9
Year 2008
Type of Item Conference Full Paper
License
Details
Bibliographic Citation I. Rexakis and M. G. Lagoudakis, “Classifier-Based Policy Representation,” in 2008 IEEE International Conference on Machine Learning and Applications (ICMLA), pp. 91–98. doi:10.1109/ICMLA.2008.31 https://doi.org/10.1109/ICMLA.2008.31
Appears in Collections

Summary

Motivated by recent proposals that view a reinforcement learning problem as a collection of classification problems, we investigate various aspects of policy representation using classifiers. In particular, we derive optimal policies for two standard reinforcement learning domains (inverted pendulum and mountain car) in both deterministic and stochastic versions and we examine their internal structure. We then proceed in an evaluation of the representational ability of a variety of classifiers for these policies, using both a multi-class and a binary formulation of the classification problem. Finally, we evaluate the actual performance of the policies learned by the classifiers in the original control problem as a function of the amount of training examples provided. Our results offer significant insight in making the reinforcement-learning-via-classification technology successfully applicable to hard learning problems.

Services

Statistics