Focus of attention in sequential decision making
- Lihong Li
- Vadim Bulitko
- Russ Greiner, Dept of Computing Science; PI of AICML
We investigate the problem of using function approximationin reinforcement learning (RL) where theagent's control policy is represented as a classifier mappingstates to actions. The innovation of this paper lieswith introducing a measure of state's decision-makingimportance. We then use an efficient approximation tothis measure as misclassification costs in learning theagent's policy. As a result, the focused learning processis shown to converge faster to better policies.
Citation
L. Li, V. Bulitko, R. Greiner. "Focus of attention in sequential decision making". Learning and Planning in Markov Processes -- Advances and Challenges, pp 43--48, July 2004.| Keywords: | MDP, machine learning, focus |
| Category: | In Workshop |
BibTeX
@misc{Li+al:L&PinMP04,
author = {Lihong Li and Vadim Bulitko and Russ Greiner},
title = {Focus of attention in sequential decision making},
Pages = {43--48},
booktitle = {Learning and Planning in Markov Processes -- Advances and
Challenges},
year = 2004,
}Last Updated: April 25, 2007Submitted by Nelson Loyola
