MatClassRSA v2 release: A MATLAB toolbox for M/EEG classification, proximity matrix construction, and visualization

Abstract

MatClassRSA is a MATLAB toolbox that performs magnetoen-cephalography and electroencephalography (M/EEG) classification and other analyses related to Representational Similarity Analysis (RSA). The toolbox is designed for cognitive neuroscience researchers who wish to perform classification-based decoding analyses of their data--often repeated trials of evoked responses--or derive Representational Dissimilarity Matrices (RDMs) as input to RSA. Organized as a collection of usercalled functions, MatClassRSA v2 comprises five main modules: Preprocessing, Reliability, Classification, RDM Computation, and Visualization. The Classification module supports multiple classifiers (LDA, RF, SVM) and classification schemes (e.g., multiclass, pairwise, cross-validated, train-test, hyperparameter optimization) and offers basic statistical analyses via permutation testing. Functions in other modules include, for example, trial-averaging and noise normalization; reliability estimation; non-classification RDM construction; and hierarchical and non-hierarchical clustering visualizations. The toolbox is freely available on GitHub under an MIT license and includes the main codebase, a User Manual, example function calls, and illustrative analyses. This preprint provides a general background of M/EEG classification for RSA as well as a narrative overview of the v2 MatClassRSA release, its updated functionalities, and illustrative analyses performed using the toolbox.