Our paper on exploring early stopping bias in the low-data regime of deep learning has been accepted at ICASSP (link)!