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

Search

Browse

My Space

Approximate policy iteration using large-margin classifiers

Lagoudakis Michael

Full record

URI: http://purl.tuc.gr/dl/dias/B95FD666-3683-44DB-8681-8CB3C2DFEC7B
Year 2003
Type of Item Conference Full Paper
License
Details
Bibliographic Citation M.G. Lagoudakis and R. Parr, “Approximate policy iteration using large-margin classifiers,” in Proceedings of the 18th International Joint Conference on Artificial Intelligence (IJCAI), 2003, pp. 1432–1434.
Appears in Collections
Conference Publications in Community School of Electrical and Computer Engineering

Summary

We present an approximate policy iteration algorithmthat uses rollouts to estimate the value of eachaction under a given policy in a subset of states anda classifier to generalize and learn the improvedpolicy over the entire state space. Using a multiclasssupport vector machine as the classifier, weobtained successful results on the inverted pendulumand the bicycle balancing and riding domains.

Available Files

Services

Statistics

 

Copyright © DIAS 2013