Training intelligent agents using LCSs in the Three-Cornered Coevolution System

In existing pattern classification systems, humans usually play a major role in creating and controlling the problem domain. In particular, humans set-up and tune the problem´s difficulty. A motivation of the work is to design and develop a system that can generate various sets of exemplars to be learned from and perform the classification tasks autonomously. The system be able to automatically adjust the problem´s difficulty based on the learners´ ability to learn. This paper introduces the Three-Cornered Coevolution System, a real implementation of the Three-Cornered Coevolution Framework proposed by Wilson, that had not been implemented previously. The system consists of three different agents that evolve to adapt with and drive the changes of the problem. Experiments show that the realised system is capable of autonomously generating various problems and triggering learning through coevolutionary process autonomously.