Вестник Нижегородского университета им. Н.И. Лобачевского. Серия: Социальные науки

16+

Title of Article

APPLICATION OF THE INFORMATION-THEORETICAL APPROACH TO MULTILAYER PERCEPTRON LEARNING IN THE PHONEME RECOGNITION PROBLEM

Issue	5
Date	2010

Section

MODELING AND SIGNAL PROCESSING METHODS

Article type	scientific article	UDC	519.234: 681.841.3
Pages	354-357	Keywords	neural networks, training, Renyi entropy, phoneme recognition

Authors

Morozov Oleg AleksandrovichNauchno-issledovatelskiy fiziko-tekhnicheskiy institut Nizhegorodskogo gosuniversiteta im. N.I. Lobachevskogo
Ovchinnikov Pavel EvgenevichNauchno-issledovatelskiy fiziko-tekhnicheskiy institut Nizhegorodskogo gosuniversiteta im. N.I. Lobachevskogo
Semin Yuriy AnatolevichNauchno-issledovatelskiy fiziko-tekhnicheskiy institut Nizhegorodskogo gosuniversiteta im. N.I. Lobachevskogo
Fidelman Vladimir RomanovichNauchno-issledovatelskiy fiziko-tekhnicheskiy institut Nizhegorodskogo gosuniversiteta im. N.I. Lobachevskogo

Abstract

A learning method based on the information theory criterion (Renyi entropy) has been proposed. Simulation results demonstrate that by using the proposed scheme the learning speed can be increased in comparison with the standard approach.

File (in Russian)