Hybrid Search Methods for Automatic Discovery of Computational Agent Schemes

This paper deals with utilizing computational agents to solve data mining tasks. A composition of agents representing hybrid computational intelligence problem solvers is proposed. Two approaches for searching the space of possible configurations are studied, namely logical reasoning and evolutionary algorithm. The dual nature of these approaches leads us to the proposition of a hybrid system able to automatically generate and verify new configurations. A simple case study is presented to show the plausibility of this approach.