Proposta de uma Classe de Perceptrons Híbridos com Aprendizagem baseada em Gradiente Descendente
Resumo
Este trabalho apresenta um modelo, chamado de dilation-erosion-linear perceptron (DELP), que consiste de uma combinação linear entre operadores não-lineares e um operador linear, para superar o dilema do passeio aleatório no problema de previsão de séries temporais financeiras. Para projetar o DELP foi apresentado um método baseado em gradiente descendente utilizando ideias do algoritmo de retropropagação do erro e uma abordagem sistemática para superar o problema da não-diferenciabilidade das operações não-lineares empregadas no modelo. Também, foi incluída uma etapa adicional para ajustar distorções de fase temporais que ocorrem na reconstrução do espaço de fase de fenômenos temporais financeiros. Por fim, uma análise experimental foi conduzida utilizando um conjunto relevante de séries financeiras, onde foram utilizadas cinco métricas e uma função de avaliação para mensurar o desempenho preditivo do modelo, e os resultados alcançados superaram aqueles obtidos por técnicas consolidadas na literatura.
Referências
de A. Araújo, R. Evolutionary learning processes to design the dilation-erosion perceptron for weather forecasting. Neural Processing Letters – Accepted For Publication.
de A. Araújo, R. (2011a). A class of hybrid morphological perceptrons with application in time series forecasting. Knowledge-Based Systems, 24(4):513–529.
de A. Araújo, R. (2011b). Translation invariant morphological time-lag added evolutionary forecasting method for stock market prediction. Expert Systems with Applications, 38(3):2835–2848.
de A. Araújo, R. (2012a). Hybrid morphological methodology for software development cost estimation. Expert Systems with Applications, 39:6129–6139.
de A. Araújo, R. (2012b). A morphological perceptron with gradient-based learning for brazilian stock market forecasting. Neural Networks, 28:61–81.
de A. Araújo, R. (2012c). A robust automatic phase-adjustment method for financial forecasting. Knowledge-Based Systems, 27:245–261.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011a). Designing dilation-erosion perceptrons with differential evolutionary learning for air pressure forecasting. In International Joint Conference on Neural Networks. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011b). Dilation-erosion perceptrons with evolutionary learning for weather forecasting. In IEEE International Conference on Systems, Man, and Cybernetics. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011c). An evolutionary approach to design dilation-erosion perceptrons for stock market indices forecasting. In ACM Genetic and Evolutionary Computation Conference. ACM.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011d). Gradient-based morphological approach for software development cost estimation. In International Joint Conference on Neural Networks. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011e). A method to overcome the random walk dilemma in financial forecasting. In International Symposium on Forecasting.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011f). Quantum-inspired method for financial time series forecasting. In IEEE Congress on Evolutionary Computation. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011g). A quantum-inspired morphological approach to solve the random walk dilemma for financial forecasting. In Congresso Brasileiro de Inteligência Computacional.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011h). Swarm-based evolutionary learning for dilation-erosion perceptrons with application in natural phenomena prediction. In IEEE Congress on Evolutionary Computation. IEEE.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. A quantum-inspired evolutionary learning process to design dilation-erosion perceptrons for financial forecasting. Learning and Nonlinear Models – Accepted For Publication.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. A swarm-based evolutionary morphological approach for binary classification problems. Learning and Nonlinear Models – Accepted For Publication.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012a). A dilation-erosion-linear perceptron for bovespa index prediction. International Conference on Intelligent Data Engineering and Automated Learning.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012b). An evolutionary morphological approach for software development cost estimation. Neural Networks, 32:285–291.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012c). A hybrid model for s&p500 index forecasting. International Conference on Artificial Neural Networks.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012d). A hybrid neuron with gradient-based learning for binary classification problems. Encontro Nacional de Inteligência Artificial ENIA.
de A. Araújo, R., Oliveira, A. L. I., and Soares, S. C. B. (2011i). A shift-invariant morphological system for software development cost estimation. Expert Systems with Applications, 38(4):4162–4168.
de A. Araújo, R. (2011a). A class of hybrid morphological perceptrons with application in time series forecasting. Knowledge-Based Systems, 24(4):513–529.
de A. Araújo, R. (2011b). Translation invariant morphological time-lag added evolutionary forecasting method for stock market prediction. Expert Systems with Applications, 38(3):2835–2848.
de A. Araújo, R. (2012a). Hybrid morphological methodology for software development cost estimation. Expert Systems with Applications, 39:6129–6139.
de A. Araújo, R. (2012b). A morphological perceptron with gradient-based learning for brazilian stock market forecasting. Neural Networks, 28:61–81.
de A. Araújo, R. (2012c). A robust automatic phase-adjustment method for financial forecasting. Knowledge-Based Systems, 27:245–261.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011a). Designing dilation-erosion perceptrons with differential evolutionary learning for air pressure forecasting. In International Joint Conference on Neural Networks. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011b). Dilation-erosion perceptrons with evolutionary learning for weather forecasting. In IEEE International Conference on Systems, Man, and Cybernetics. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011c). An evolutionary approach to design dilation-erosion perceptrons for stock market indices forecasting. In ACM Genetic and Evolutionary Computation Conference. ACM.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011d). Gradient-based morphological approach for software development cost estimation. In International Joint Conference on Neural Networks. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011e). A method to overcome the random walk dilemma in financial forecasting. In International Symposium on Forecasting.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011f). Quantum-inspired method for financial time series forecasting. In IEEE Congress on Evolutionary Computation. IEEE.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011g). A quantum-inspired morphological approach to solve the random walk dilemma for financial forecasting. In Congresso Brasileiro de Inteligência Computacional.
de A. Araújo, R., de Oliveira, A. L. I., Soares, S. C. B., and de L. Meira, S. R. (2011h). Swarm-based evolutionary learning for dilation-erosion perceptrons with application in natural phenomena prediction. In IEEE Congress on Evolutionary Computation. IEEE.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. A quantum-inspired evolutionary learning process to design dilation-erosion perceptrons for financial forecasting. Learning and Nonlinear Models – Accepted For Publication.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. A swarm-based evolutionary morphological approach for binary classification problems. Learning and Nonlinear Models – Accepted For Publication.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012a). A dilation-erosion-linear perceptron for bovespa index prediction. International Conference on Intelligent Data Engineering and Automated Learning.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012b). An evolutionary morphological approach for software development cost estimation. Neural Networks, 32:285–291.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012c). A hybrid model for s&p500 index forecasting. International Conference on Artificial Neural Networks.
de A. Araújo, R., Oliveira, A. L. I., and Meira, S. (2012d). A hybrid neuron with gradient-based learning for binary classification problems. Encontro Nacional de Inteligência Artificial ENIA.
de A. Araújo, R., Oliveira, A. L. I., and Soares, S. C. B. (2011i). A shift-invariant morphological system for software development cost estimation. Expert Systems with Applications, 38(4):4162–4168.
Publicado
23/07/2013
Como Citar
ARAÚJO, Ricardo de A.; MEIRA, Silvio; OLIVEIRA, Adriano L. I..
Proposta de uma Classe de Perceptrons Híbridos com Aprendizagem baseada em Gradiente Descendente. In: CONCURSO DE TESES E DISSERTAÇÕES (CTD), 26. , 2013, Maceió/AL.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2013
.
p. 41-46.
ISSN 2763-8820.
