作者: P. Gallinari , S. Thiria , F. Badran , F. Fogelman-Soulie
DOI: 10.1016/0893-6080(91)90071-C
关键词: Set (abstract data type) 、 Artificial neural network 、 Generic property 、 Extension (predicate logic) 、 Linear discriminant analysis 、 Computer science 、 Nonlinear system 、 Perceptron 、 Algorithm
摘要: Abstract We study the relations between discriminant analysis and multilayer perceptrons used for classification tasks. first consider linear networks prove formal equivalence two techniques in this case. then present a set of experiments on problems with increasing degree nonlinearity. This allows to extension result nonlinear nets investigate data transformations successive layers these nets. Finally, we show evidence generic properties MLPs classifiers.