ASSESSMENT OF LONGITUDINAL DISPERSION COEFFICIENT BY MEANS OF DIFFERENT NEURAL NETWORKS

摘要

Multi-Layer Perceptron, Fuzzy and Radial-Basis Function neural networks, Nearest Neighbour approach, linear regression, 2~(nd) order curvilinear regression, and 'classical' empirical formulae have been applied for evaluation of longitudinal dispersion coefficient for a river reach. In general, results achieved by means of each type of neural networks outperforms these obtained by other techniques.