Amplifying pathological detection in EEG signaling pathways through cross-dataset transfer learning

Mohammad-Javad Darvishi Bayazi, Mohammad Sajjad Ghaemi, Timothée Lesort, Md Rifat Arefin, Jocelyn Faubert, Irina Rish

Abstract: Highlights•Using large datasets improves performance on tasks with limited labeled data.•Data and model scaling techniques improve pathology classification accuracy.•Small, generic models (e.g., ShallowNet) perform well on single datasets.•Larger models excel in transfer learning, particularly with diverse datasets.•Width is more important than depth in neural networks for pathology detection.