Article 11220

Title of the article

THE APPLICATION OF SINGLE-LAYER PERCEPTRON FOR SOLVING PATTERN CLASSIFICATION TASK OF ELECTRONIC COMPONENTS IN ORDER TO IMPROVE THE QUALITY AND RELIABILITY OF ONBOARD EQUIPMENT 

Authors

Mishanov Roman Olegovich, candidate of technical sciences, assistant, sub-department of design and technology electronic systems and devices, Samara University (34 Moskovskoe highway, Samara, Russia), E-mail: kipres@ssau.ru 

Index UDK

621.382 

DOI

10.21685/2307-4205-2020-2-11 

Abstract

The article is devoted to the possibility of singlelayer perceptron application with a different number of hidden neurons for classifying electronic components into classes of acceptable and potentially defective instances. The forecasted (output) and informative (input) parameters of the integrated chips are used as the initial data. Network training was carried out using the software «Deductor Academic », which allowed determining the main network parameters: the values of synaptic weights and thresholds for each neuron. After training each network, the responses to each example of training sample were determined and compared with reference values. Evaluation of the network efficiency was carried out according to the accuracy of modeling and classification. 

Key words

single-layer perceptron, individual forecasting, classification, electronic component, linearly separable classes, pattern, hidden neuron, neuron network, neuron network training 

Download PDF

 

Дата создания: 17.07.2020 11:16
Дата обновления: 17.07.2020 14:08