Optimization of Neocognitron Parameters for Recognition of Symbols in State Registration Signs
Published: 10.10.2012
Authors: Kangin D.N., Kurov A.V.
Published in issue: #1(1)/2012
DOI: 10.18698/2308-6033-2012-1-30
Category: Information technology
The neuron net Fucusima Neocognitron is investigated. The paper describes methods defining the threshold values, level measurements, core parameters of Gauss function used for calculating output level values. Ways of optimizing the parameters are proposed. The results obtained show that the coefficient of recognizing test selection is 97%.