Ensemble learning based on functional connectivity and Riemannian geometry for robust workload estimation

International audienceContext Passive Brain-Computer Interface (pBCI) has recently gained in popularity through its applications, e.g. workload and attention assessment. Nevertheless, one of the main limitations remains the important intra-and inter-subject variability. We propose a robust approach relying on ensemble learning, grounded in functional connectivity and Riemannian geometry to mitigate the high variability of the data with a large and diverse panel of classifiers