scientific article; zbMATH DE number 1107551

From MaRDI portal
Publication:4373480





We concentrate on three widely-used neural network paradigms. The perspective adopted could be described as numerically-oriented. The role neural network algorithms play relative to cognitive science, for instance, is not considered. The perspective adopted is also critical: we are interested in current or potential applications of neural networks that are important vis-à-vis cluster or discriminant analysis or closely related methods. The ``dream of computational emulation of biological or other physical activities is only occasionally referenced. Section 2 will review the multilayer perceptron, viewed primarily as a method of nonlinear discrimination or regression. Section 3 will review so-called ``self-organized, topological mappings. Among objectives of these techniques are nonlinear dimensionality reduction and cluster analysis. Section 4 will review the Hopfield network and its use as a tool for combinatorial optimization. Section 5 will present a short literature review of direct linkages between conventional multivariate data analysis methods and neural network approaches. For the entire collection see [Zbl 0836.00014].




Cited in
(32)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4373480)