This repository contains pytorch code for training and testing self-certifying classifiers. It implements linearized deep assignment flows operating on features extracted by a variant of ResNet18.

For details on implemented methods, please consult the corresponding paper on ArXiv.

The code was developed by the IPA group at Heidelberg University.