To Learn or Not to Learn

Anupam Joshi
Department of Computer Sciences
Purdue University
West Lafayette, IN 47907-1398
USA
Phone: 317-494-7821, Fax: 317-494-0739
email: joshi@cs.purdue.edu

Abstract

Multiagent systems in which agents interact with each other are
now being proposed as a solution to many problems which can be
grouped together under the ``distributed problem solving''
umbrella. For such systems to work properly, it is necessary that
agents learn from their environment and adapt their behaviour
accordingly. In this paper we present a system which uses a
combination of neuro--fuzzy learning and static adaptation to
coordinate the activity of multiple agents. An epistemic utility
based formulation is used to automatically generate the exemplars
for learning, making the process unsupervised. The system has
been developed in the context of a scientific computing scenario.