TY - JOUR

T1 - Calculation of Analogs of Lyapunov Exponents for Different Types of EEG Time Series using Artificial Neural Networks

AU - Chernykh, German A.

AU - Dmitrieva, Ludmila A.

AU - Kuperin, Yuri A.

PY - 2019

Y1 - 2019

N2 - The relevance of the study is due to the need to develop reliable methods of classification of relativelyshort and complex time series, where known methods do not always give an acceptable result. The study aims atdeveloping a method for constructing new numerical characteristics of time series, which are analogs of Lyapunovexponents for complex dynamic systems in their physical and mathematical sense.The developed method is new because it is based on the use of committees of artificial neural networks. The methodwas tested on time series of electroencephalograms, in particular, for comparative analysis of electroencephalogramsof subjects in a state of meditation with their background records of electroencephalograms. Statistically significantresults are obtained, on the basis of which, on the one hand, it can be concluded that the developed technique can beeffectively applied to the time series of electroencephalograms. On the other hand, the developed technique allowsdrawing certain conclusions about the specifics of synchronization of large ensembles of neurons for the subjectspracticing meditation. The developed methods and the results are of theoretical interest for researchers involved inthe analysis of time series. Also, these methods and results are of practical value for neurophysiologists who studythe processes of meditation by quantitative methods.

AB - The relevance of the study is due to the need to develop reliable methods of classification of relativelyshort and complex time series, where known methods do not always give an acceptable result. The study aims atdeveloping a method for constructing new numerical characteristics of time series, which are analogs of Lyapunovexponents for complex dynamic systems in their physical and mathematical sense.The developed method is new because it is based on the use of committees of artificial neural networks. The methodwas tested on time series of electroencephalograms, in particular, for comparative analysis of electroencephalogramsof subjects in a state of meditation with their background records of electroencephalograms. Statistically significantresults are obtained, on the basis of which, on the one hand, it can be concluded that the developed technique can beeffectively applied to the time series of electroencephalograms. On the other hand, the developed technique allowsdrawing certain conclusions about the specifics of synchronization of large ensembles of neurons for the subjectspracticing meditation. The developed methods and the results are of theoretical interest for researchers involved inthe analysis of time series. Also, these methods and results are of practical value for neurophysiologists who studythe processes of meditation by quantitative methods.

KW - Lyapunov exponents

KW - electroencephalograms

KW - embedding in lag space

KW - artificial neural networks

M3 - Article

VL - 108

SP - 2553

EP - 2569

JO - Ekoloji

JF - Ekoloji

SN - 1300-1361

IS - 28

ER -