A comparative study of neural network based feature extraction paradigms

The projection maps and derived classification accuracies of a neural netwo rk (NN) implementation of Sammon's mapping, an auto-associative NN (AANN) a nd a multilayer perceptron (MLP) feature extractor are compared with those of the conventional principal component analysis (PCA). Tested on five real -world databases, the MLP provides the highest classification accuracy at t he cost of deforming the data structure, whereas the linear models preserve the structure but usually with inferior accuracy. (C) 1999 Elsevier Scienc e B.V. All rights reserved.