2006 | OriginalPaper | Buchkapitel
Memetic Algorithms
verfasst von : Natalio Krasnogor, Alberto Aragón, Joaquín Pacheco
Erschienen in: Metaheuristic Procedures for Training Neutral Networks
Verlag: Springer US
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
This chapter introduces and analyzes a memetic algorithm approach for the training of artificial neural networks, more specifically multilayer perceptrons. Our memetic algorithm is proposed as an alternative to gradient search methods, such as
backpropagation
, which have shown limitations when dealing with rugged landscapes with many poor local optimae. The aim of our work is to design a training strategy that is able to cope with difficult error manyfolds, and to quickly deliver trained neural networks that produce small errors. A method such as the one we proposed might also be used as an “online” training strategy.