On the Use of Quasi-Newton-Based Training of a Feedforward Neural Network for Time Series Forecasting

Abstract

This article examines the effectiveness of using a quasi-Newton-based training of a feedforward neural network for forecasting. We have developed a novel quasi-Newton-based training algorithm using a generalised logistic function. We have shown that a well-designed feedforward structure can lead to a good forecast without the use of the more complicated feedback/feedforward structure of the recurrent network.