On the Recurrent Neural Network Induction By Evolutionary Computations With Applications in Forecasting

Ion Ciuca, Elena Jitaru

Abstract

This paper considers recurrent neural networks with one hidden layer. The recurrence is realized in the hidden layer by inner connections. These neural structures are very suitable for dynamic process neurocontrol. The paper presents an algorithm from evolutionary computation for evolving the parameters and structure of these neural networks. Also, by way of example, a simulation application in forecasting, based on time-series of a dynamic process is presented.

Keywords

evolving, induction, evolutionary programming, neural networks

View full article