Slice Oriented Tensor Decomposition of EEG Data for Feature Extraction in Space, Frequency and Time Domains

Zhao, Qibin; Caiafa, César F.; Cichocki, Andrzej; Zhang, Liqing; Phan, Anh Huy

doi:10.1007/978-3-642-10677-4_25

Cited by 12 publications

(11 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To our knowledge this is the first iterative algorithm to build a heterogeneous model. Previous cases have used alternating algorithms [10,28] where the ranks of each type of decomposition are chosen a priori.…”

Section: Resultsmentioning

confidence: 99%

A greedy algorithm for model selection of tensor decompositions

Brockmeier

Prı́ncipe

Phan

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

Various tensor decompositions use different arrangements of factors to explain multi-way data. Components from different decompositions can vary in the number of parameters. Allowing a model to contain components from different decompositions results in a combinatoric number of possible models. We consider model selection to balance approximation error with the number of parameters, but due to the number of possibilities post-hoc model selection is infeasible. Instead, we incrementally build a model. This approach is analogous to sparse coding with a union of dictionaries. The proposed greedy approach can estimate a model consisting of a combination of tensor decompositions.

show abstract

Section: Resultsmentioning

confidence: 99%

A greedy algorithm for model selection of tensor decompositions

Brockmeier

Prı́ncipe

Phan

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Some of the tensor decomposition formulations for a 3 mode tensor shown in Figure are as follows: (a) parallel factor analysis (PARAFAC), equivalently known as canonical decomposition (CANDECOMP)/canonical polyadic decomposition (CPD), where a tensor is described as a sum of rank-1 tensors, (b) Tucker decomposition, where the tensor is expressed as the sum of outer products of different rank factor matrices in each mode weighted by a hypercube, and more recently, (c) slice oriented decomposition, where a tensor is represented as the sum of an outer product of column vector of the n th mode factor matrix with tensor hyper-slices of the remaining modes. This decomposition, implemented on a three-way tensor of electroencephalogram (EEG) data, is seen to be robust to outliers and captures patterns in the slices …”

Section: Introductionmentioning

confidence: 99%

Structure-Preserving Joint Non-negative Tensor Factorization to Identify Reaction Pathways Using Bayesian Networks

Puliyanda

Sivaramakrishnan

et al. 2021

J. Chem. Inf. Model.

View full text Add to dashboard Cite

Extracting meaningful information from spectroscopic data is key to species identification as a first step to monitoring chemical reactions in unknown complex mixtures. Spectroscopic data obtained over multiple process modes (temperature, residence time) from different sensors [Fourier transform infrared (FTIR), proton nuclear magnetic resonance ( 1 H NMR)] comprise hidden complementary information of the underlying chemical system. This work proposes an approach to jointly capture these hidden patterns in a structure-preserving and interpretable manner using coupled non-negative tensor factorization to achieve uniqueness in decomposition. Projections onto the modes of spectral channels, specific to each sensor, are interpreted as pseudo-component spectra, while projections onto the shared process modes are interpreted as the corresponding pseudo-component concentrations across temperature and residence times. Causal structure inference among these pseudo-component spectra (using Bayesian networks) is then used to identify plausible reaction pathways among the identified species representing each pseudo-component. Tensor decomposition of the FTIR data enables the development of reaction sequences based on the identified functional groups, while that of 1 H NMR by itself is lacking in mechanism development as it solely reveals the proton environments in a pseudocomponent. However, jointly parsing spectra from both the sensors is seen to capture complementary information, wherein insights into the proton environment from 1 H NMR disambiguate pseudo-components that have similar FTIR peaks. A scalable method of parallelizing tensor decomposition to handle high-dimensional modes in process data by using grid tensor factorization, while being robust to process data artifacts like outliers, noise, and missing data, has been developed.

show abstract

“…The reason is that the EEG signal is an electric activity of the brain measured by electrodes (channels), which placed on the scalp is a superposition of electric signals which are produced by a synchronous activity of numerous neurons. [4] The spatial propagation of the electric signal in the complex brain tissue is far from being straightforward. Moreover, various groups of neurons participate in various and sometimes independent tasks.…”

Section: Introductionmentioning

confidence: 99%

A Data Driven Approach for Resting-state EEG signal Classification of Schizophrenia with Control Participants using Random Matrix Theory

Liu¹,

Zhang²,

Ye³

et al. 2017

Preprint

View full text Add to dashboard Cite

Resting state electroencephalogram (EEG) abnormalities in clinically high-risk individuals (CHR), clinically stable first-episode patients with schizophrenia (FES), healthy controls (HC) suggest alterations in neural oscillatory activity. However, few studies directly compare these anomalies among each types. Therefore, this study investigated whether these electrophysiological characteristics differentiate clinical populations from one another, and from non-psychiatric controls. To address this question, resting EEG power and coherence were assessed in 40 clinically high-risk individuals (CHR), 40 first-episode patients with schizophrenia (FES), and 40 healthy controls (HC). These findings suggest that resting EEG can be a sensitive measure for differentiating between clinical disorders.This paper proposes a novel data-driven supervised learning method to obtain identification of the patients mental status in schizophrenia research. According to Marchenko-Pastur Law, the distribution of the eigenvalues of EEG data is divided into signal subspace and noise subspace. A test statistic named LES that embodies the characteristics of all eigenvalues is adopted. different classifier and different feature(LES test function) are selected for experiments, we have shown that using von Neumann Entropy as LES test function combine with SVM classifier could obtain the best average classification accuracy during three classification among HC, FES and CHR of Schizophrenia group with EEG signal. It is worth noting that the result of LES feature extraction with the highest classification accuracy is around 90% in two classification(HC compare with FES) and around 70% in three classification. Where the classification accuracy higher than 70% could be used to assist clinical diagnosis.

show abstract

Slice Oriented Tensor Decomposition of EEG Data for Feature Extraction in Space, Frequency and Time Domains

Cited by 12 publications

References 12 publications

A greedy algorithm for model selection of tensor decompositions

A greedy algorithm for model selection of tensor decompositions

Structure-Preserving Joint Non-negative Tensor Factorization to Identify Reaction Pathways Using Bayesian Networks

A Data Driven Approach for Resting-state EEG signal Classification of Schizophrenia with Control Participants using Random Matrix Theory

Contact Info

Product

Resources

About