Learning from Inconsistent EEG Sensors with Differentiable Channel Reordering

Aaqib Saeed
David Grangier
Olivier Pietquin
Neil Zeghidour
ICASSP 2021 (to appear)


We propose CHARM, a method for training a single architecture across inconsistent input channels. Our work is motivated by Electroencephalography (EEG) where data collection protocols from different headsets result in varying channel ordering and number, which limits the feasibility of transferring trained systems across datasets. Our approach builds upon attention mechanisms to estimate a latent permutation matrix from each input signal, and map input channels to a canonical order. \Ours{} is differentiable and can be composed further with architectures expecting consistent channel ordering to build end-to-end trainable classifiers. We perform experiments on four EEG classification datasets and demonstrate the efficacy of \Ours{} via simulated reordering and masking on input channels. Moreover, our method improves transfer of pre-trained representations between datasets collected with different protocols.