Cultivating Desired Behaviour: Policy Teaching Via Environment-Dynamics Tweaks

Rabinovich, Z., Dufton, L., Larson, K. and Jennings, N. (2010) Cultivating Desired Behaviour: Policy Teaching Via Environment-Dynamics Tweaks. In: The 9th International Conference on Autonomous Agents and Multiagent Systems.

Download

Preview

Published Version
PDF
157Kb

Abstract

In this paper we study, for the first time explicitly, the
implications of endowing an interested party (i.e. a teacher) with the
ability to modify the underlying $\emph{dynamics}$ of the environment,
in order to encourage an agent to learn to follow a specific
policy. We introduce a cost function which can be used by the teacher
to balance the modifications it makes to the underlying environment
dynamics, with the learner's performance compared to some ideal,
desired, policy. We formulate teacher's problem of determining optimal
environment changes as a planning and control problem, and empirically
validate the effectiveness of our model.

Creators:	Zinovi Rabinovich, Lachlan Dufton, Kate Larson, Nick Jennings
Item Type:	Conference or Workshop Item
Keywords:	Teacher-learner, control theory, Kullback-Leibler Rate
Research Group:	Intelligence, Agents, Multimedia
Deposited On:	05 Feb 2010 13:09 by Rabinovich, Zinovi
ID Code:	18470
Last Modified:	18 Feb 2010 16:28