The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
No reconhecimento de padrões usando redes neurais, é muito difícil para pesquisadores ou usuários projetarem uma arquitetura de rede neural ideal para uma tarefa específica. É possível que qualquer tipo de arquitetura de rede neural obtenha uma certa medida de taxa de reconhecimento. É, no entanto, difícil obter analiticamente uma arquitetura de rede neural ideal para uma tarefa específica na relação de reconhecimento e eficácia do treinamento. Neste artigo, é proposto um método evolutivo de treinamento e projeto de redes neurais feedforward. No método proposto, uma rede neural é definida como um indivíduo e redes neurais cujas arquiteturas são iguais às de uma espécie. Essas redes são avaliadas pelo MSE (Mean Square Error) normalizado que apresenta o desempenho de uma rede para padrões de treinamento. Então, suas arquiteturas evoluem de acordo com uma regra de evolução aqui proposta. Arquiteturas de redes neurais, ou seja, espécies, são avaliadas por outra medida de critérios comparada com os critérios dos indivíduos. Os critérios avaliam o indivíduo mais superior da espécie e a velocidade de evolução da espécie. As espécies aumentam ou diminuem em tamanho populacional de acordo com os critérios. A regra da evolução gera arquiteturas de rede neural um pouco diferentes das espécies superiores. O método proposto, portanto, pode gerar diversas arquiteturas de redes neurais. O projeto e treinamento de redes neurais que executam 3 simples
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Hiroki TAKAHASHI, Masayuki NAKAJIMA, "Evolutional Design and Training Algorithm for Feedforward Neural Networks" in IEICE TRANSACTIONS on Information,
vol. E82-D, no. 10, pp. 1384-1392, October 1999, doi: .
Abstract: In pattern recognition using neural networks, it is very difficult for researchers or users to design optimal neural network architecture for a specific task. It is possible for any kinds of neural network architectures to obtain a certain measure of recognition ratio. It is, however, difficult to get an optimal neural network architecture for a specific task analytically in the recognition ratio and effectiveness of training. In this paper, an evolutional method of training and designing feedforward neural networks is proposed. In the proposed method, a neural network is defined as one individual and neural networks whose architectures are same as one species. These networks are evaluated by normalized M. S. E. (Mean Square Error) which presents a performance of a network for training patterns. Then, their architectures evolve according to an evolution rule proposed here. Architectures of neural networks, in other words, species, are evaluated by another measurement of criteria compared with the criteria of individuals. The criteria assess the most superior individual in the species and the speed of evolution of the species. The species are increased or decreased in population size according to the criteria. The evolution rule generates a little bit different architectures of neural network from superior species. The proposed method, therefore, can generate variety of architectures of neural networks. The designing and training neural networks which performs simple 3
URL: https://global.ieice.org/en_transactions/information/10.1587/e82-d_10_1384/_p
Copiar
@ARTICLE{e82-d_10_1384,
author={Hiroki TAKAHASHI, Masayuki NAKAJIMA, },
journal={IEICE TRANSACTIONS on Information},
title={Evolutional Design and Training Algorithm for Feedforward Neural Networks},
year={1999},
volume={E82-D},
number={10},
pages={1384-1392},
abstract={In pattern recognition using neural networks, it is very difficult for researchers or users to design optimal neural network architecture for a specific task. It is possible for any kinds of neural network architectures to obtain a certain measure of recognition ratio. It is, however, difficult to get an optimal neural network architecture for a specific task analytically in the recognition ratio and effectiveness of training. In this paper, an evolutional method of training and designing feedforward neural networks is proposed. In the proposed method, a neural network is defined as one individual and neural networks whose architectures are same as one species. These networks are evaluated by normalized M. S. E. (Mean Square Error) which presents a performance of a network for training patterns. Then, their architectures evolve according to an evolution rule proposed here. Architectures of neural networks, in other words, species, are evaluated by another measurement of criteria compared with the criteria of individuals. The criteria assess the most superior individual in the species and the speed of evolution of the species. The species are increased or decreased in population size according to the criteria. The evolution rule generates a little bit different architectures of neural network from superior species. The proposed method, therefore, can generate variety of architectures of neural networks. The designing and training neural networks which performs simple 3
keywords={},
doi={},
ISSN={},
month={October},}
Copiar
TY - JOUR
TI - Evolutional Design and Training Algorithm for Feedforward Neural Networks
T2 - IEICE TRANSACTIONS on Information
SP - 1384
EP - 1392
AU - Hiroki TAKAHASHI
AU - Masayuki NAKAJIMA
PY - 1999
DO -
JO - IEICE TRANSACTIONS on Information
SN -
VL - E82-D
IS - 10
JA - IEICE TRANSACTIONS on Information
Y1 - October 1999
AB - In pattern recognition using neural networks, it is very difficult for researchers or users to design optimal neural network architecture for a specific task. It is possible for any kinds of neural network architectures to obtain a certain measure of recognition ratio. It is, however, difficult to get an optimal neural network architecture for a specific task analytically in the recognition ratio and effectiveness of training. In this paper, an evolutional method of training and designing feedforward neural networks is proposed. In the proposed method, a neural network is defined as one individual and neural networks whose architectures are same as one species. These networks are evaluated by normalized M. S. E. (Mean Square Error) which presents a performance of a network for training patterns. Then, their architectures evolve according to an evolution rule proposed here. Architectures of neural networks, in other words, species, are evaluated by another measurement of criteria compared with the criteria of individuals. The criteria assess the most superior individual in the species and the speed of evolution of the species. The species are increased or decreased in population size according to the criteria. The evolution rule generates a little bit different architectures of neural network from superior species. The proposed method, therefore, can generate variety of architectures of neural networks. The designing and training neural networks which performs simple 3
ER -