Hyperborders in the Voronoi-Diabram-Based Neural Net for Pattern Classification

Published

May 20, 2002

Author(s)

Camillo A. Gentile, M Sznaier

Abstract

We propose a neural network to answer a point query in Rn partitioned based on the Voronoi diagram. Our novel design offers the potential to reduce both the number of neurons and connection weights of previous designs, employing a cost function which enables a tradeoff between the two to suit a specific implementation. Our simplified structure requires neither delay weights nor complex neurons, while retaining the main advantage of previous designs to furnish precise values for the neurons and connection weights, as opposed to trial and error iterations or ad-hoc parameters.

Citation

International Joint Conference on Neural Networks

Pub Type

Journals

Download Paper

Local Download

Keywords

Neural networks, Pattern classification, Voronoi diagrams

Citation

Gentile, C. and Sznaier, M. (2002), Hyperborders in the Voronoi-Diabram-Based Neural Net for Pattern Classification, International Joint Conference on Neural Networks, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=151100 (Accessed October 15, 2025)