A Non-Linear Tensor Tracking Algorithm for Analysis of Incomplete Multi-Channel EEG Data

Tensor decomposition is a popular tool to analyse and process data which can be represented by a higher-order tensor structure. In this paper, we consider tensor tracking in challenging situations where the observed data are streaming and incomplete. Specifically, we proposed a non-linear formulatio...

Full description

Saved in:
Bibliographic Details
Published in2018 12th International Symposium on Medical Information and Communication Technology (ISMICT) pp. 1 - 6
Main Authors Linh-Trung, Nguyen, Minh-Chinh, Truong, Nguyen, Viet-Dung, Abed-Meraim, Karim
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.03.2018
Subjects
Adaptation models
Cost function
Electrodes
Electroencephalography
Matrix decomposition
Time-frequency analysis
Online AccessGet full text

Cover

More Information
Summary:Tensor decomposition is a popular tool to analyse and process data which can be represented by a higher-order tensor structure. In this paper, we consider tensor tracking in challenging situations where the observed data are streaming and incomplete. Specifically, we proposed a non-linear formulation of the PETRELS cost function and based on which we proposed NL-PETRELS subspace and tensor tracking algorithms. The non-linear function allows us to improve the convergence rate. We also illustrated the use of our proposed tensor tracking for incomplete multi-channel electroencephalogram (EEG) data in a real-life experiment in which the data can be represented by a third-order tensor.
ISSN:2326-8301
DOI:10.1109/ISMICT.2018.8573711