Published in CVPR-workshops, 2025
Despite plenty of research in the last 10 years, we have made limited progress towards generating robust machine learning models. Therefore, in this paper we explore the question of whether current datasets are large enough to train robust image classifiers.
Published in CVPR, 2024
A large scale comparison of methods of creating 1-Lipschitz convolutions, considering both theoretical properties as well as experimental results.
Published in arXiv, 2023
We show a shortcoming with currently popular activation functions in 1-Lipschitz networks and propose an activation function that provably overcomes this limitation.
Published in ECCV, 2022
We introduce an efficient rescaling-based layer that allows us to train state-of-the-art certifiably robust image classifiers.