Determination of Features of the Dependence of Neuron Net Output Coordinates on Other Coordinates and Parameters of Neuron Net

Ptitchkin, V.

Поиск по всему репозиторию:

Открыть/скачать файлы документа

40-45.pdf (376.7Kb)

Автор

Ptitchkin, V.

Дата издания

1999

Издательство

BPI

Показать полное описание

Библиографическое описание

Ptitchkin, V. Determination of Features of the Dependence of Neuron Net Output Coordinates on Other Coordinates and Parameters of Neuron Net / V. Ptitchkin // International Conference on Neural Networks and Artificial Intelligence ICNNAI'99 = Международная конференция по нейронным сетям и искусственному интеллекту ICNNAI'99 : Proceedings, Brest, Belarus, 12–15 October 1999 / Brest Polytechnic Institute, Department of Computers and Laboratory of Artificial Neural Networks, Belarus Special Interest Group of International Neural NetWork Society, International Neural NetWork Society, Belarusian State University of Informatics and Radioelectronics (Belarus), Belarusian Academy of Sciences, Institute of Engineering Cybemetics (Belarus), Universidad Politechnica de Valencia (Spain), Institute of Computer Information Technologies (Ukraine, Ternopil) ; ed. V. Golovko. – Brest : BPI, 1999. – Р. 40–45.

Аннотация

The paper presents the deduction of main relationships of back-propagation algorithm on the basis of approximation of the unknown dependence of the efficiency factor on neural net parameters by the linear part of Taylor series. Such an approach for deduction of known results is intended to relax the limits of application of the back-propagation algorithm and to eliminate some of its shortcomings on the basis of new methods of approximation of the before-mentioned dependence. As an example, this paper examines the possibility of application of back-propagation algorithm in the case of non-differentiable activation functions.