Open AccessAutomatic Generation of Neural Networks
Author(s) -
A. Fiszelew,
Paola Verónica Britos,
G. Perichisky,
Ramón García-Martínez
Publication year - 2009
Publication title -
revista eletrônica de sistemas de informação
Language(s) - English
Resource type - Journals
ISSN - 1677-3071
DOI - 10.21529/resi.2003.0201001
Subject(s) - computer science , artificial neural network , artificial intelligence , machine learning , evolutionary algorithm , domain (mathematical analysis) , process (computing) , genetic algorithm , backpropagation , mathematics , mathematical analysis , operating system
This work deals with methods for finding optimal neural network architectures to learn par-ticular problems. A genetic algorithm is used to discover suitable domain specific architectures; this evolutionary algorithm applies direct codification and uses the error from the trained network as a per-formance measure to guide the evolution. The network training is accomplished by the back-propagation algorithm; techniques such as training repetition, early stopping and complex regulation are employed to improve the evolutionary process results. The evaluation criteria are based on learn-ing skills and classification accuracy of generated architectures