Título: Otimização da generalização de uma Rede Neural utilizando Algoritmos Genéticos
Autores: Ventura, Erico S.; Amorim, Cláudio A.
Resumo: A method to maximize the generalization capability of a Multilayer Perceptron (MLP) is purposed, and applied to the recognition of printed characters. A Genetic Algorithm is used to find an optimal match between the fonts in the training set and the number of hidden nodes in the MLP. The results obtained were satisfactory, and reinforce the idea that the number of hidden nodes of a neural network should vary in accordance with the specific data used to train it.
Palavras-chave: Artificial Neural Networks; Genetic Algorithms; Generalization; Optimization; Character recognition
Páginas: 5
Código DOI: 10.21528/CBRN2005-107
Artigo em PDF: CBRN2005_107.pdf
Arquivo BibTex: CBRN2005_107.bib