Role Model of Search in Agents' Parameter-Space

Abstract

In this article we elaborate the formal model of roles in computational multi-agent systems (CMAS) in description logic. The CMAS model is enriched by a role-based model representing search (e.g. hill-climbing, genetic algorithms) in general search space. The choice of solution representation is important for successful and quick finding of the optimal solution. We apply the search model to optimization in the parameter space of data mining methods and employ it in a meta-learning scenario.