Implemented and trained various deep learning models, including Convolutional Neural Networks (CNN), Convolutional Recurrent Neural Networks (CRNN), and Convolutional Transformer (Conformer) models, tailored for EEG signal classification.
Utilized the BCI Competition IV Dataset 2a for model training and validation, achieving test classification accuracies that exceeded the traditional benchmark of 69.43% by at least 1.5%.
Implemented nontrivial data augmentation and a novel CNN architecture to reach a peak test accuracy of 74.97% over 50 epochs.
Technologies: Python, PyTorch

results

results