The discrete wavelet neural network and its application in oscillographic chronopotentiometric determination

The structure and algorithm of the discrete wavelet neural network (DWNN) a re described. The network is constructed by the error back propagation neur al network using Morlet mother wavelet basic function as node activation fu nction. The effect of wavelet base number, learning rate factor and momentu m factor on prediction are discussed. The experimental results of the quant itative computation for the concentration of mono-component and multi-compo nent in oscillographic chronopotentiometric determination (OCPD) show that number of epochs is less than 1000, the recovery is between 94.37% and 104. 3%. Compared with standard back propagation neural network, DWNN has higher convergence rate and prediction accuracy. (C) 2001 Elsevier Science B.V. A ll rights reserved.