Riemannian distances and divergences

Introduction

This code accompanies a the publication "Assessment of Riemannian distances and divergences for SSVEP-based BCI" which aims at assessing the impact of several distances/divergences on a real EEG dataset.

Important files

• main.m

  Classifies SSVEP trials from 12 subjects, using MDM with distances/divergences described in the aforementioned publication: arithmetic, harmonic, Riemannian, log-euclid, Kullback-Leibler, S-divergence, $\alpha$-divergence, Bhattacharrya, Wasserstein, and Jeffreys. The main output of this file are the classification accuracies and the computation time for each method/metric

• alpha_cross_validation.m

  In the $\alpha$-divergence, the value of $\alpha$ is deternined through cross-validation.

• loaddata.m

  Implement a function called in main.m. The path to the dataset is hardcoded herein.

• swelling_effect_analysis.m

  Computes the determinants and traces of means of covariance matrices computed with different distances/divergences. It also computes them for each individual covariance matrix used in the computation of the means.

  Only 1 subject and one class are used for illustrative purposes.