Optimización en la elaboración de redes neuronales artifciales adaptativas usando una metodología de algoritmo de poda
##plugins.themes.bootstrap3.article.sidebar##
Publicado:
Dez 1, 2017
##plugins.themes.bootstrap3.article.main##
Resumo
Las redes neuronales artifciales feedforward y multicapa (RNA-MFF) han demostrado ser de gran alcance en la aproximación de funciones; sin embargo, su aplicación en problemas reales a menudo se limita a la experimentación del usuario, ya que la elección de arquitectura adecuada es un proceso que requiere conocimiento y experiencia. En este artículo se demuestra la capacidad de adaptación de la metodología del algoritmo de poda para encontrar el número óptimo de neuronas en la capa oculta de una RNA-MFF. La metodología se probó en dos conjuntos diferentes de problemas de referencia: modelación de la resistencia y del asentamiento, en hormigón de alto desempeño. Ambos conjuntos se utilizaron para analizar el tamaño de la arquitectura inicial de la red neuronal artifcial y para asegurar que el número superior propuesto de neuronas ocultas se pueda evitar en exceso.AbstractFeedforward and multi-layer artifcial neural networks (RNA-MFF) have been shown to be powerful in the approximation of functions; however, its application to real problems is often limited to user experimentation, since choosing the right architecture is a process that requires knowledge and experience. This article demonstrates the adaptability of the pruning algorithm methodology in fnding the optimal number of neurons in the hidden layer of an RNA-MFF. The methodology was tested on two different sets of reference problems: resistance and settling modeling, in high performance concrete. Both sets were used to analyze the initial architecture size of the artifcial neural network and to ensure that the proposed higher number of hidden neurons can be avoided in excess.Resumo As redes neurais artifciais avançadas e multicamadas feedforward (RNA-MFF) têm demostrado ter grande alcance na aproximação de funções; porém, a aplicação em problemas reais geralmente é limitada à experimentação do usuário, dado que escolher a arquitetura adequada é um processo que precisa conhecimento e experiência. Neste artigo foi demonstrada a adaptabilidade da metodologia do algoritmo de poda para encontrar o número ótimo de neurônios na camada oculta de uma RNA-MFF. A metodologia foi testada em dois conjuntos diferentes de problemas de referência: modelagem de resistência e sedimentação, em concreto de alto desempenho. Ambos os conjuntos foram utilizados para analisar o tamanho da arquitetura inicial da rede neural artifcial e garantir que o maior número de neurônios ocultos propostos possa ser evitado em excesso.
Downloads
Não há dados estatísticos.
##plugins.themes.bootstrap3.article.details##
Como Citar
González-Salcedo, L. O., Gotay-Sardiñas, J., Roodschild, M., Ernesto-Will, A. L., & Rodríguez, S. (2017). Optimización en la elaboración de redes neuronales artifciales adaptativas usando una metodología de algoritmo de poda. Ingenio Magno, 8(1), 44-56. Recuperado de http://revistas.ustatunja.edu.co/index.php/ingeniomagno/article/view/1388
Seção
Artículos Vol. 8-1
DECLARACIÓN DE ORIGINALIDAD DE ARTÍCULO PRESENTADO
Por medio del presente documento, certifico(amos) que el artículo que se presenta para posible publicación en la revista institucional INGENIO MAGNO del Centro de Investigaciones de Ingeniería Alberto Magno CIIAM de la Universidad Santo Tomás, seccional Tunja, es de mi (nuestra) entera autoría, siendo su contenido producto de mi (nuestra) directa contribución intelectual y aporte al conocimiento.
Todos los datos y referencias a publicaciones hechas están debidamente identificados con su respectiva nota bibliográfica y en las citas que se destacan como tal. De requerir alguna clase de ajuste o corrección, comunicaré(emos) de tal procedimiento con antelación a los responsables de la revista.
Por lo anteriormente expresado, declaro(amos) que el material presentado en su totalidad se encuentra conforme a la legislación aplicable en materia de propiedad intelectual e industrial de ser el caso, y por lo tanto, me(nos) hago (hacemos) absolutamente responsable(s) de cualquier reclamación relacionada a esta.
En caso que el artículo presentado sea publicado, manifiesto(amos) que cedo(emos) plenamente a la Universidad Santo Tomás, seccional Tunja, los derechos de reproducción del mismo.
Referências
Aschen, C. (1982). Interpreting and using regression. Newbury Park: Sage University Paper.
Ascombe, T. (1973). Graphs and Statistical Analysis. The American Statistician, 27(1), 17-21.
Asuncion, A. y Newman, D. (2007). UCI Machine Learning Repository. Recuperado de http://www.ics.uci.edu/~mlearn/MLRepository.html
Bentz, D., Peltz, M. y Winpigler, J. (2009). Early-age properties of cement-based materials: II. Influence of water-to-cement ratio. ASCE Journal of Materials in Civil Engineering, 21(9), 512-517.
Bhadeshia, H., Dimitriu, R., Forsik, S., Pak, J. y Ryu, J. (2009). Performance of neural networks in materials science. Materials Science and Technology, 25(4), 504-510.
Castellano, G., Fanelli, A. y Pelillo, P. (1997). An iterative pruning algorithm for feedforward neural networks. IEEE Transactions on Neural
Networks, 8(3), 519-531.
Chaouachi, A., Kamel, R., Ichikawa, R., Hayashi, H. y Nagasaka, K.
(2009). Neural network ensemble-based solar power generation short-term forecasting. International Journal of Engineering and Mathematical Sciences, 3(6), 40-46.
Constantiner, D. y Dackzo, J. (2002). Not all applications are created equal: Selecting the appropriate SCC performance targets. First North American Conference on the Design and Use of Self Consolidating Concrete. Chicago.
Cybenko, G. (1989). Approximation by superpositions of a sigmoidal function. Mathematics of Control, Sigma and Systems, 2(4), 303-314.
Díaz, W., Will, A. y González, S. (2011). Un método para estimación de calidad de datos en redes neuronales y aplicación a problemas de hormigón de alto performance. Mecánica Computacional, 30(43), 3381-3381.
Doukim, C., Dargham, J. y Chekima, A. (2010). Finding the number of hidden neurons for a MLP neural networks using coarse to fne search technique. Recuperado de https://www.semanticscholar.org/paper/Finding-the-number-of-hidden-neurons-for-an-MLP-neDoukim-Dargham/c442c3ac2589e1f048d-
5d5827556763c5b7fbdfa
González, S., Guerrero, Z., Delvasto, A. y Will, A. (2012). Red neuronal Artifcial para estimar la resistencia a compresión, en concretos fbroreforzados con acero. Ciencia e Ingeniería Neogranadina, 22(1), 19-41.
González, S. (2014). Diseño de mezclas de concreto reforzado con fbras metálicas y de polipropileno, usando inteligencia artifcial (tesis doctoral). Cali: Universidad del Valle.
González, S., Guerrero, Z., Delvasto, A. y Will, A. (2014). Estimación del Índice de Tenacidad Flexural I5 en concretos fbroreforzados, usando redes neuronales artifciales. Revista Colombiana de Materiales, 5, 24-29.
Hashemi-Nejad, M. y Fekri, H. (2009). Shortterm wind speed prediction using adaptive neural networks. Europe Wind Energy Conference and Exhibition 2009. Marsella, Francia.
Hornik, K., Stinchcombe, M. y White, H. (1989). Multilayer feedforward networks are universal approximators. Neural Networks, 2(5), 359-366.
Huang, R. y Shurong, T. (2009). Evolving product unit neural networks with particle swarm optimization. Proceedings of the 2009 Fifth International Conference on Image and Graphics.
Jinchuan, K. y Xinzhe, L. (2008). Empirical analysis of optimal hidden neurons in neural networks modeling for stock prediction. Proceedings of the Pacifc-Asia Workshop on Computational Intelligence and Industrial Applications, 40, 828-832.
Kiranyaz, S., Ince, T., Yildirim, A. y Gabbouj, M. (2009). Evolutionary artifcial neural networks by multi-dimensional particle swarm optimization. Neural Networks, 22(10), 1448-1462.
Koza, J. y Rice, J. (1991). Genetic generation of both the weigths and arquitecture for neural networks. Proceedings of IEEE, IJCNN, Seattle
(WA), pp. 397-404.
Kwok, T. y Yeung, D. (1997). Constructive algorithms for structure learning in feedforward neural networks for regression problems. IEEE Transactions on Neural Networks, 8(3), 630-645.
MacKay, D. (1992). A practical bayesian framework for backpropagation networks. Neural Computation, 4(3), 448-472.
Martínez, E. (2005). Errores frecuentes en la interpretación del coefciente de determinación lineal. Anuario Jurídico y Económico Escurialense, 38, 315-332.
Panchal, G., Ganatra, A., Kosta, Y. y Panchal, D. (2011). Behaviour analysis of multilayer perceptrons with multiple hidden neurons and
hidden layers. International Journal of Computer Theory and Engineering, 3(2), 332-337.
Prechelt, L. (1998). Automatic early stopping using cross validation: quantifying the criteria. Neural Networks, 11(4), 761-767.
Reed, R. (1993). Pruning algorithms-a survey. Transations of Neural Networks, 4(5), 740-747.
Sánchez, D. (2000). Tecnología del Concreto y Mortero. Bogotá: Bhandar.
Sharma, S. y Chandra, P. (2010a). Constructive neural networks: A review. International Journal of Engineering Science and Technology, 2(12), 7847-7855.
Sharma, S. y Chandra, P. (2010b). An adaptive slope sigmoidal function cascading neural networks algorithm. Proceedings of the IEEE, Third International Conference on Emerging Trends in Engineering and Technology.
Sharma, S. y Chandra, P. (2010c). An adaptive slope basic dynamic node creation algorithm for single hidden layer neural networks. Proceedings of the IEEE, International Conference on Computational Intelligence and Communication Systems.
Subirats, J., Franco, L. y Jerez, J. (2012). C-MANTEC: A novel constructive neural network algorithm incorporating competition between neurons. Neural Networks, 26, 130-140.
Tompkins, C. (1992). Using and interpreting linear regression and correlation analyses: Some cautions and considerations. Clinical
Aphasiology Paper, 21, 35-46.
Wei, G. (2008). Evolutionary neural network based on new ant colony algorithm. International Symposium on Computational Intelligence and Design IEEE.
Yao, X. y Liu, Y. (1997). A new revolutionary system for evolving artificial neural networks. Transactions on Neural Networks, 8(3), 694-713.
Yeh, I. (1998). Modeling of strength of high-performance concrete using artifcial neural networks. Cement and Concrete Research, 28(12), 1797-1808.
Yeh, I. (2007). Modeling slump flow of concrete using second-order regressions and artifcial neural networks. Cement and Concrete Composites, 29(6), 474-480.
Yuan, H., Xiong, F. y Huai, X. (2003). A method for estimating the number of hidden neurons in feed-forward neural networks based on information entropy. Computers and Electronics in Agriculture, 40(1-3), 57-64.
Ascombe, T. (1973). Graphs and Statistical Analysis. The American Statistician, 27(1), 17-21.
Asuncion, A. y Newman, D. (2007). UCI Machine Learning Repository. Recuperado de http://www.ics.uci.edu/~mlearn/MLRepository.html
Bentz, D., Peltz, M. y Winpigler, J. (2009). Early-age properties of cement-based materials: II. Influence of water-to-cement ratio. ASCE Journal of Materials in Civil Engineering, 21(9), 512-517.
Bhadeshia, H., Dimitriu, R., Forsik, S., Pak, J. y Ryu, J. (2009). Performance of neural networks in materials science. Materials Science and Technology, 25(4), 504-510.
Castellano, G., Fanelli, A. y Pelillo, P. (1997). An iterative pruning algorithm for feedforward neural networks. IEEE Transactions on Neural
Networks, 8(3), 519-531.
Chaouachi, A., Kamel, R., Ichikawa, R., Hayashi, H. y Nagasaka, K.
(2009). Neural network ensemble-based solar power generation short-term forecasting. International Journal of Engineering and Mathematical Sciences, 3(6), 40-46.
Constantiner, D. y Dackzo, J. (2002). Not all applications are created equal: Selecting the appropriate SCC performance targets. First North American Conference on the Design and Use of Self Consolidating Concrete. Chicago.
Cybenko, G. (1989). Approximation by superpositions of a sigmoidal function. Mathematics of Control, Sigma and Systems, 2(4), 303-314.
Díaz, W., Will, A. y González, S. (2011). Un método para estimación de calidad de datos en redes neuronales y aplicación a problemas de hormigón de alto performance. Mecánica Computacional, 30(43), 3381-3381.
Doukim, C., Dargham, J. y Chekima, A. (2010). Finding the number of hidden neurons for a MLP neural networks using coarse to fne search technique. Recuperado de https://www.semanticscholar.org/paper/Finding-the-number-of-hidden-neurons-for-an-MLP-neDoukim-Dargham/c442c3ac2589e1f048d-
5d5827556763c5b7fbdfa
González, S., Guerrero, Z., Delvasto, A. y Will, A. (2012). Red neuronal Artifcial para estimar la resistencia a compresión, en concretos fbroreforzados con acero. Ciencia e Ingeniería Neogranadina, 22(1), 19-41.
González, S. (2014). Diseño de mezclas de concreto reforzado con fbras metálicas y de polipropileno, usando inteligencia artifcial (tesis doctoral). Cali: Universidad del Valle.
González, S., Guerrero, Z., Delvasto, A. y Will, A. (2014). Estimación del Índice de Tenacidad Flexural I5 en concretos fbroreforzados, usando redes neuronales artifciales. Revista Colombiana de Materiales, 5, 24-29.
Hashemi-Nejad, M. y Fekri, H. (2009). Shortterm wind speed prediction using adaptive neural networks. Europe Wind Energy Conference and Exhibition 2009. Marsella, Francia.
Hornik, K., Stinchcombe, M. y White, H. (1989). Multilayer feedforward networks are universal approximators. Neural Networks, 2(5), 359-366.
Huang, R. y Shurong, T. (2009). Evolving product unit neural networks with particle swarm optimization. Proceedings of the 2009 Fifth International Conference on Image and Graphics.
Jinchuan, K. y Xinzhe, L. (2008). Empirical analysis of optimal hidden neurons in neural networks modeling for stock prediction. Proceedings of the Pacifc-Asia Workshop on Computational Intelligence and Industrial Applications, 40, 828-832.
Kiranyaz, S., Ince, T., Yildirim, A. y Gabbouj, M. (2009). Evolutionary artifcial neural networks by multi-dimensional particle swarm optimization. Neural Networks, 22(10), 1448-1462.
Koza, J. y Rice, J. (1991). Genetic generation of both the weigths and arquitecture for neural networks. Proceedings of IEEE, IJCNN, Seattle
(WA), pp. 397-404.
Kwok, T. y Yeung, D. (1997). Constructive algorithms for structure learning in feedforward neural networks for regression problems. IEEE Transactions on Neural Networks, 8(3), 630-645.
MacKay, D. (1992). A practical bayesian framework for backpropagation networks. Neural Computation, 4(3), 448-472.
Martínez, E. (2005). Errores frecuentes en la interpretación del coefciente de determinación lineal. Anuario Jurídico y Económico Escurialense, 38, 315-332.
Panchal, G., Ganatra, A., Kosta, Y. y Panchal, D. (2011). Behaviour analysis of multilayer perceptrons with multiple hidden neurons and
hidden layers. International Journal of Computer Theory and Engineering, 3(2), 332-337.
Prechelt, L. (1998). Automatic early stopping using cross validation: quantifying the criteria. Neural Networks, 11(4), 761-767.
Reed, R. (1993). Pruning algorithms-a survey. Transations of Neural Networks, 4(5), 740-747.
Sánchez, D. (2000). Tecnología del Concreto y Mortero. Bogotá: Bhandar.
Sharma, S. y Chandra, P. (2010a). Constructive neural networks: A review. International Journal of Engineering Science and Technology, 2(12), 7847-7855.
Sharma, S. y Chandra, P. (2010b). An adaptive slope sigmoidal function cascading neural networks algorithm. Proceedings of the IEEE, Third International Conference on Emerging Trends in Engineering and Technology.
Sharma, S. y Chandra, P. (2010c). An adaptive slope basic dynamic node creation algorithm for single hidden layer neural networks. Proceedings of the IEEE, International Conference on Computational Intelligence and Communication Systems.
Subirats, J., Franco, L. y Jerez, J. (2012). C-MANTEC: A novel constructive neural network algorithm incorporating competition between neurons. Neural Networks, 26, 130-140.
Tompkins, C. (1992). Using and interpreting linear regression and correlation analyses: Some cautions and considerations. Clinical
Aphasiology Paper, 21, 35-46.
Wei, G. (2008). Evolutionary neural network based on new ant colony algorithm. International Symposium on Computational Intelligence and Design IEEE.
Yao, X. y Liu, Y. (1997). A new revolutionary system for evolving artificial neural networks. Transactions on Neural Networks, 8(3), 694-713.
Yeh, I. (1998). Modeling of strength of high-performance concrete using artifcial neural networks. Cement and Concrete Research, 28(12), 1797-1808.
Yeh, I. (2007). Modeling slump flow of concrete using second-order regressions and artifcial neural networks. Cement and Concrete Composites, 29(6), 474-480.
Yuan, H., Xiong, F. y Huai, X. (2003). A method for estimating the number of hidden neurons in feed-forward neural networks based on information entropy. Computers and Electronics in Agriculture, 40(1-3), 57-64.