Тип публикации: доклад, тезисы доклада, статья из сборника материалов конференций
Конференция: International Scientific Multi-Conference on Industrial Engineering and Modern Technologies (FarEastCon); Vladivostok, RUSSIA; Vladivostok, RUSSIA
Год издания: 2018
Ключевые слова: pattern recognition, nonparametric systems, high-dimension, decomposition of training sample
Аннотация: From positions of the principles of training selection decomposition and collective estimation the synthesis technique of multilevel nonparametric systems of pattern recognition for the multialternate classification problem is offered. Their application provides high computing performance of information processing of big dimension. Two approaches are considered. Poorly dependent feature sets of the classified objects are in case of the former used. Considering the assumption of independence of feature sets the generalized decisive rule of maximum likelihood is under construction. The basis of the second method is made by a dichotomy method. At each its stage we form the family of the private decision functions corresponding to various feature sets of the classified objects with the subsequent their integration in the non-linear decisive rule by means of methods of nonparametric statistics. At the same time formation of the generalized decision on situation belonging to this or that class is carried out in space of values of private decision functions. The offered technique allows to use technology of parallel calculations.
Издание
Журнал: 2018 INTERNATIONAL SCIENTIFIC MULTI-CONFERENCE ON INDUSTRIAL ENGINEERING AND MODERN TECHNOLOGIES (FAREASTCON)
Место издания: NEW YORK
Издатель: IEEE
Персоны
- Lapko Alexandr V. (RAS, SB, Inst Comp Modelling, Reshetnev Siberian State Univ Sci & Technol, Krasnoyarsk, Russia)
- Lapko Vasily A. (RAS, SB, Inst Comp Modelling, Reshetnev Siberian State Univ Sci & Technol, Krasnoyarsk, Russia)
- Yuronen Ekaterina A. (Siberian Fed Univ, Reshetnev Siberian State Univ Sci & Technol, Krasnoyarsk, RussiaProceedings Paper)
Вхождение в базы данных
Информация о публикациях загружается с сайта службы поддержки публикационной активности СФУ. Сообщите, если заметили неточности.