Toward Aristotelian Medical Representations: Backpropagation-Free Layer-wise Analysis for Interpretable Generalized Metric Learning on MedMNIST
arXiv:2604.06017v1 Announce Type: new
Abstract: While deep learning has achieved remarkable success in medical imaging, the “black-box” nature of backpropagation-based models remains a significant barrier to clinical adoption. To bridge this gap, we p…