Adaptive learning with covariate shift-detection for motor imagery-based brain–computer interface

Haider Raza, Hubert Cecotti, Y Li, Girijesh Prasad

Research output: Contribution to journalArticle

24 Citations (Scopus)

Abstract

A common assumption in traditional supervised learning is the similar probability distribution of data betweenthe training phase and the testing/operating phase. When transitioning from the training to testing phase, a shift in the probability distribution of input data is known as a covariate shift. Covariate shifts commonly arise in a wide range of real-world systems such as electroencephalogram-based brain–computer interfaces (BCIs). In such systems, there is a necessity for continuous monitoring of the process behavior, and tracking the state of the covariate shifts to decide about initiating adaptation in a timely manner. This paper presents a covariate shift-detection and -adaptation methodology, and its application to motor imagery-based BCIs. A covariateshift-detection test based on an exponential weighted moving average model is used to detect the covariate shift in the features extracted from motor imagery-based brain responses. Following the covariate shift-detection test, the methodology initiates an adaptation by updating the classifier during the testing/operating phase. The usefulness of the proposed method is evaluated using real-world BCI datasets (i.e. BCI competition IV dataset 2A and 2B). The results show a statistically significant improvement in the classification accuracy of the BCI system over traditional learning and semi-supervised learning methods.
LanguageEnglish
Pages3085-3096
Number of pages11
JournalSoft Computing
Volume20
Issue number8
DOIs
Publication statusPublished - 28 Nov 2015

Fingerprint

Brain computer interface
Adaptive Learning
Covariates
Supervised learning
Probability distributions
Testing
Probability Distribution
Moving Average Model
Electroencephalography
Semi-supervised Learning
Methodology
Weighted Average
Supervised Learning
Brain
Classifiers
Updating
Imagery
Classifier
Monitoring
Range of data

Keywords

  • Adaptive learning ·
  • Transductive learning
  • · Covariate shift-detection ·
  • Brain–computer interfaces

Cite this

@article{ddddcaefbdb84c6f948852925fcec266,
title = "Adaptive learning with covariate shift-detection for motor imagery-based brain–computer interface",
abstract = "A common assumption in traditional supervised learning is the similar probability distribution of data betweenthe training phase and the testing/operating phase. When transitioning from the training to testing phase, a shift in the probability distribution of input data is known as a covariate shift. Covariate shifts commonly arise in a wide range of real-world systems such as electroencephalogram-based brain–computer interfaces (BCIs). In such systems, there is a necessity for continuous monitoring of the process behavior, and tracking the state of the covariate shifts to decide about initiating adaptation in a timely manner. This paper presents a covariate shift-detection and -adaptation methodology, and its application to motor imagery-based BCIs. A covariateshift-detection test based on an exponential weighted moving average model is used to detect the covariate shift in the features extracted from motor imagery-based brain responses. Following the covariate shift-detection test, the methodology initiates an adaptation by updating the classifier during the testing/operating phase. The usefulness of the proposed method is evaluated using real-world BCI datasets (i.e. BCI competition IV dataset 2A and 2B). The results show a statistically significant improvement in the classification accuracy of the BCI system over traditional learning and semi-supervised learning methods.",
keywords = "Adaptive learning ·, Transductive learning, · Covariate shift-detection ·, Brain–computer interfaces",
author = "Haider Raza and Hubert Cecotti and Y Li and Girijesh Prasad",
year = "2015",
month = "11",
day = "28",
doi = "10.1007/s00500-015-1937-5",
language = "English",
volume = "20",
pages = "3085--3096",
journal = "Soft Computing",
issn = "1432-7643",
number = "8",

}

Adaptive learning with covariate shift-detection for motor imagery-based brain–computer interface. / Raza, Haider; Cecotti, Hubert; Li, Y; Prasad, Girijesh.

In: Soft Computing, Vol. 20, No. 8, 28.11.2015, p. 3085-3096.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Adaptive learning with covariate shift-detection for motor imagery-based brain–computer interface

AU - Raza, Haider

AU - Cecotti, Hubert

AU - Li, Y

AU - Prasad, Girijesh

PY - 2015/11/28

Y1 - 2015/11/28

N2 - A common assumption in traditional supervised learning is the similar probability distribution of data betweenthe training phase and the testing/operating phase. When transitioning from the training to testing phase, a shift in the probability distribution of input data is known as a covariate shift. Covariate shifts commonly arise in a wide range of real-world systems such as electroencephalogram-based brain–computer interfaces (BCIs). In such systems, there is a necessity for continuous monitoring of the process behavior, and tracking the state of the covariate shifts to decide about initiating adaptation in a timely manner. This paper presents a covariate shift-detection and -adaptation methodology, and its application to motor imagery-based BCIs. A covariateshift-detection test based on an exponential weighted moving average model is used to detect the covariate shift in the features extracted from motor imagery-based brain responses. Following the covariate shift-detection test, the methodology initiates an adaptation by updating the classifier during the testing/operating phase. The usefulness of the proposed method is evaluated using real-world BCI datasets (i.e. BCI competition IV dataset 2A and 2B). The results show a statistically significant improvement in the classification accuracy of the BCI system over traditional learning and semi-supervised learning methods.

AB - A common assumption in traditional supervised learning is the similar probability distribution of data betweenthe training phase and the testing/operating phase. When transitioning from the training to testing phase, a shift in the probability distribution of input data is known as a covariate shift. Covariate shifts commonly arise in a wide range of real-world systems such as electroencephalogram-based brain–computer interfaces (BCIs). In such systems, there is a necessity for continuous monitoring of the process behavior, and tracking the state of the covariate shifts to decide about initiating adaptation in a timely manner. This paper presents a covariate shift-detection and -adaptation methodology, and its application to motor imagery-based BCIs. A covariateshift-detection test based on an exponential weighted moving average model is used to detect the covariate shift in the features extracted from motor imagery-based brain responses. Following the covariate shift-detection test, the methodology initiates an adaptation by updating the classifier during the testing/operating phase. The usefulness of the proposed method is evaluated using real-world BCI datasets (i.e. BCI competition IV dataset 2A and 2B). The results show a statistically significant improvement in the classification accuracy of the BCI system over traditional learning and semi-supervised learning methods.

KW - Adaptive learning ·

KW - Transductive learning

KW - · Covariate shift-detection ·

KW - Brain–computer interfaces

U2 - 10.1007/s00500-015-1937-5

DO - 10.1007/s00500-015-1937-5

M3 - Article

VL - 20

SP - 3085

EP - 3096

JO - Soft Computing

T2 - Soft Computing

JF - Soft Computing

SN - 1432-7643

IS - 8

ER -