The CILT multi-agent learning system

Abstract

A multi-agent learning system consists of two or more learners or learning agents that co-exist in a learning environment. The learners have the ability to learn together, thus utilizing one another's strengths and alleviating individual weaknesses. This paper describes the cooperative inductive learning team (CILT) multi-agent learning system that consists of two or more learners which induce rules from training examples. The learners cooperate by sharing their knowledge with one another. In this way, the individual results are improved and a team knowledge-base, that contains the best team results, is created.