Matlab/Octave Machine Learning Toolbox for linear classification with applications in Brain-Computer Interfaces (BCI). This toolbox is distributed with GPL license along a tutorial chapter Machine learning for BCI in the book.

• Linear models

  • Linear Discriminant Analysis (LDA)
  • Support Vector Machine (SVM)
  • Ridge Regression (RR)

• Validation strategies

  • Hold-out
  • Random sampling
  • K-fold cross validation
  • Leave-one-out bootstrap

• Performance measures (classification and regression)

  • Accuracy (ACC)
  • Area Under the ROC curve (AUC)
  • Cohen's Kappa (k)
  • Means Square Error (MSE)
  • Correlation coefficient (corr)

• Demo Datasets

  • Motor Imagery (MI) with CSP features
  • P300 Speller with temporal features
  • ECoG finger movement prediction dataset

• Download current version here
• Extract in a subfolder
• Add path to Matlab/Octave with "addpath(genpath('path-to-mltool'))" to execute any function in the toolbox

All functions in the toolbox contain detailled documentation with parameters definition.

• classifier folder contains all the linear models estimation functions (only binary classification)
• performance folder contains all the performance computation functions
• validation folder contains all the pvalidation loops
• figures folder contains the script used for generating figures in the chapter

• MI Dataset is extracted from dataset IIa, BCI competition IVand has been pre-precessed by Fabien Lotte.
• P300 Dataset is extracted from subject APM of UAM P300 Dataset using feature extraction described in here
• Finger flexion ECog dataset is extracted from dataset 4 from BCI competition IV

• Linear SVM Solver Copyright 2006 Olivier Chapelle
• The computation of te Area under the ROC curve is performed using svmroccurve.m that has been extracted from SVM-KM

• Rémi Flamary
• Alain Rakotomamonjy