Open AccessModel-free classification of multivariate time-series based on epsilon-complexity theory.
Author(s) -
B. S. Darkhovsky,
Ras Control,
Yuri A. Dubnov,
A. Yu. Popkov
Publication year - 2020
Publication title -
trudy kolʹskogo naučnogo centra ran
Language(s) - English
Resource type - Journals
ISSN - 2307-5252
DOI - 10.37614/2307-5252.2020.8.11.023
Subject(s) - multivariate statistics , series (stratigraphy) , binary number , binary classification , mathematics , feature (linguistics) , space (punctuation) , feature vector , algorithm , computer science , pattern recognition (psychology) , artificial intelligence , statistics , arithmetic , support vector machine , paleontology , linguistics , philosophy , biology , operating system
This work is devoted to a new model-free approach to a problem of binary classification of multivariate time-series. The approach is based on the original theory of epsilon-complexity which allows almost every mapping that satisfies Hoelder condition, be characterized by a pair of real numbers –complexity coefficients. Thus we can form a feature space in which a classification problem can be formulated and solved. We provide an example of classification of real EEG signals.