The Application of Cascade Correlation Neural Networks to River Flow Forecasting

This chapter provides an insight into the background and the motivation that led to the development of cascade correlation. The mechanics of the algorithm are described in some detail, as is the quickprop (Falman, 1988) update rule that is often used in conjunction with cascade correlation. This is followed by a number of suggested modifications that may render the algorithm more suitable for regression tasks such as river flow forecasting. Finally there is a flow-routing case study using the River Trent in the UK.