Rethinking Deep Learning in fNIRS for Abnormal Data Exclusion

The author argues that current deep learning models in fNIRS struggle to exclude abnormal data, proposing a simple yet effective method using metric learning and supervised techniques to enhance reliability.