Model and Implementation of Self-Organising Neural Network for Searching Discovery in Databases

Abstract

This paper presents selected results of possibilities of building model of self-organising neural network using SPHINX language and NEURONIX Toolbox. By implementation used to algorithm Winner Take All (WTA) and Winner Take Most (WTM). For practical verification worked up example contains base date consist information about parameters of printers working in each departments of office. Databases contain knowledge related with characteristics of technical parameters each printer and their exploitation conditions. Neural Network used to discovery regularities in information consists of 20 input and 20 neurones. Examples contain about 200 training vectors. For implementation elaborated some functions such as init (initialisation first-weight of neurone), zmien_wagi (change of weight of neurone), zapisz_wagi (write weights all neurones), algorythm_WTA (algorithm WTA used to Euclidean distance), etc. Moreover implemented two functions for searching information in databases, one function for count of numbers printers, which must be to buy, etc. Also wrote two functions responsible for visualisations discoveries. Key words: self-organising neural network, searching discovery in databases