ACM logo

The ACM Journal of Experimental Algorithmics


Volume 7, Article 12, 2002


Relational Concept Learning by Cooperative Evolution

by

Filippo Neri

http://www.jea.acm.org/2002/NeriRelational/

Abstract:

Concept learning is a computationally demanding task that involves searching large hypothesis spaces containing candidate descriptions. Stochastic search combined with parallel processing provides a promising approach to dealing with such a task. Learning systems based on distributed genetic algorithms (GAs) have proved able to find concept descriptions as accurate as those found by other state-of-the-art learning systems, although the GA approach has the drawback of being more computationally expensive. We show here how a suitable architectural design, cooperative evolution, allows the solution of complex applications with GA-based learning systems while keeping the computational cost acceptable, thanks to the effective exploitation of distributed computation. A variety of experimental settings is used and an explanation of the experimental findings is proposed.

Keywords: distributed genetic algorithm, first order logic

Categories: I.2.6 (Computing Methodologies) Learning: Relational Concept Learning; Induction: GA-Based Learning System

Received
Accepted
Final Revision
Published




Last updated and validated May 11, 2002, by editor@jea.acm.org