This repository contains different DL model implementations for EEG-based classification of mental states (attention vs. inattention to continuous auditory input).
EEGnet... notebooks implement CNN-based binary classifiers that attempt to predict whether the subject was attending to the auditory input or mind-wandering.
...stim_reconst notebooks implement a stimulus reconstruction approach, in which a convolutional neural net minimises the MSE loss between the ground truth auditory stimulus and its reconstruction from EEG signal.
Alternative architectures are also present, including capsule networks (CapsNET notebooks), and Siamese networks (Siamese_L1... notebooks).