Title :
The neural network of linear approximation
Author :
Kozynchenko, Vladimir A. ; Prus, Anna I.
Author_Institution :
St. Petersburg State Univ., St. Petersburg, Russia
fDate :
June 30 2014-July 4 2014
Abstract :
In this paper we consider a three-layer neural network, carrying out a polygonal approximation of the training set of data. In the process of supervised learning, neural network divides the input training set of data into n-dimensional simplex. We construct a hyperplane approximating output training signals for each simplex. This paper presents algorithm of learning a neural network. Neural network can be used to solve problems of prediction, recognition and classification of images.
Keywords :
approximation theory; learning (artificial intelligence); neural nets; hyperplane; image classification; image prediction problem; image recognition; input training data set; linear approximation; n-dimensional simplex; neural network learning; output training signal approximation; polygonal approximation; supervised learning; three-layer neural network; Biological neural networks; Educational institutions; Linear approximation; Neurons; Piecewise linear approximation; Training;
Conference_Titel :
Computer Technologies in Physical and Engineering Applications (ICCTPEA), 2014 International Conference on
Conference_Location :
St. Petersburg
Print_ISBN :
978-1-4799-5315-8
DOI :
10.1109/ICCTPEA.2014.6893289
