Feature learning for information-extreme classifier
No Thumbnail Available
Date
2017
Journal Title
Journal ISSN
Volume Title
Publisher
Sumy State University
Theses
Date of Defense
Scientific Director
Speciality
Date of Presentation
Abstract
The feature learning algorithm for
information-extreme classifier by clustering of Fast
Retina Keypoint binary descriptor, calculated for
local features, and usage of spatial pyramid kernel for
increasing noise immunity and informativeness of
feature representation are considered. Proposed a
method of parameters optimization for feature
extractor and decision rules based on multi-level
coarse features coding using information criterion and
population-based search algorithm.
Keywords
coarse coding, Fast Retina Keypoint, feature extraction, machine learning, classifier, information criterion, машинное обучение, классификатор, информационный критерий, машинне навчання, класифікатор, информаційний критерій
Citation
Korobov, A.
Feature learning for information-extreme classifier [Текст] / A. Korobov, A. Moskalenko, V. Moskalenko // Advanced Information Systems and Technologies : proceedings of the V international scientific conference, Sumy, May 17-19 2017/ Edited by S.І. Protsenko, V.V. Shendryk. - Sumy : Sumy State University, 2017. - P. 92-94.