FAST AUTOMATIC CONFIGURATION OF ARTIFICIAL NEURAL NETWORKS USED FOR BINARY PATTERNS RECOGNITION

Abstract

This paper prcsents a powcrful method of an automatically generated architccturc of ncural networks uscd for binary pattems rccognition, which can quickly and automatically rcducc synapses in a way of minimally reducing a quality of rccognition and a quality of generalization. Moreovcr, this method computes all weights in two runs over a leaming sequence, what makes this method vcry fast. First, the method calculatcs all binary featureś for each pattem and then weights are computcd. Furthcrmore, there is a quality of generalization considered because it is one o f the most important factors o f recognition whilc using neural networks.