Using an artificial neural network to approximate the temporal evolution function of the lorenz system

• Main Authors: Andrea Martiniano, Ricardo Pinto Ferreira, Arthur Ferreira, Aleister Ferreira, Renato José Sassi
• Format: Article
• Language: Portuguese
• Published: Centro Federal de Educação Tecnológica Celso Suckow da Fonseca 2016-04-01
• Series: Revista Produção e Desenvolvimento
• Subjects: Rede Neural Artificial; Sistema de Lorentz; Aproximação de Função.
• Online Access: http://revistas.cefet-rj.br/index.php/producaoedesenvolvimento/article/view/94

The main objective of this paper is to approximate the temporal evolution function of the Lorenz system using Artificial Neural Networks (ANN) type Multilayer Perceptron (MLP). Apart from this main objective, as a specific objective, presents the basic concepts of ANN, a brief history of chaos theory and the Lorentz system. The methodology used in the structuring of this paper was defined as bibliographic and experimental. Currently, there is great interest in models of neural networks to solve unconventional and complex problems, in this context the ANN have emerged as an alternative for numerous applications in various areas of knowledge. The results of the experiments indicate positively to the use of ANN. It is hoped that this paper encourage the use of ANN in complex applications where learning, association, generalization and abstraction are needed to support decision-making. It was concluded that the use of ANN could be an alternative for solving problems involving approximation functions.