The largest EEG-based BCI reproducibility study for open science: the MOABB benchmark

arXiv:2404.15319v1 Announce Type: cross
Abstract: Objective. This study conduct an extensive Brain-computer interfaces (BCI) reproducibility analysis on open electroencephalography datasets, aiming to assess existing solutions and establish open and reproducible benchmarks for effective comparison within the field. The need for such benchmark lies in the rapid industrial progress that has given rise to undisclosed proprietary solutions. Furthermore, the scientific literature is dense, often featuring challenging-to-reproduce evaluations, making comparisons between existing approaches arduous.
Approach. Within an open framework, 30 machine learning …

abstract analysis arxiv bci benchmark benchmarks brain comparison computer cs.ai cs.hc cs.lg datasets eeg eess.sp industrial interfaces lies open science progress q-bio.nc reproducibility science solutions study type