no code implementations • 22 Apr 2024 • Leon Bungert, Tim Laux, Kerrek Stinson
We connect adversarial training for binary classification to a geometric evolution equation for the decision boundary.
1 code implementation • 30 May 2023 • Leon Bungert, Nicolás García Trillos, Matt Jacobs, Daniel Mckenzie, Đorđe Nikolić, Qingsong Wang
Although deep neural networks have achieved super-human performance on many classification tasks, they often exhibit a worrying lack of robustness towards adversarially generated examples.
no code implementations • 28 Nov 2022 • Leon Bungert, Kerrek Stinson
In this paper we prove Gamma-convergence of a nonlocal perimeter of Minkowski type to a local anisotropic perimeter.
no code implementations • 19 May 2022 • Leo Schwinn, Leon Bungert, An Nguyen, René Raab, Falk Pulsmeyer, Doina Precup, Björn Eskofier, Dario Zanca
The reliability of neural networks is essential for their use in safety-critical applications.
no code implementations • 26 Nov 2021 • Leon Bungert, Nicolás García Trillos, Ryan Murray
We establish an equivalence between a family of adversarial training problems for non-parametric binary classification and a family of regularized risk minimization problems where the regularizer is a nonlocal perimeter functional.
1 code implementation • 24 Nov 2021 • Leon Bungert, Jeff Calder, Tim Roith
In this work we prove uniform convergence rates for solutions of the graph infinity Laplace equation as the number of vertices grows to infinity.
1 code implementation • 4 Jun 2021 • Leon Bungert, Tim Roith, Daniel Tenbrinck, Martin Burger
We propose a novel strategy for Neural Architecture Search (NAS) based on Bregman iterations.
1 code implementation • 10 May 2021 • Leon Bungert, Tim Roith, Daniel Tenbrinck, Martin Burger
In contrast to established methods for sparse training the proposed family of algorithms constitutes a regrowth strategy for neural networks that is solely optimization-based without additional heuristics.
Ranked #164 on Image Classification on CIFAR-10
1 code implementation • 23 Mar 2021 • Leon Bungert, René Raab, Tim Roith, Leo Schwinn, Daniel Tenbrinck
Despite the large success of deep neural networks (DNN) in recent years, most neural networks still lack mathematical guarantees in terms of stability.
1 code implementation • 24 Feb 2021 • Leo Schwinn, An Nguyen, René Raab, Leon Bungert, Daniel Tenbrinck, Dario Zanca, Martin Burger, Bjoern Eskofier
The susceptibility of deep neural networks to untrustworthy predictions, including out-of-distribution (OOD) data and adversarial examples, still prevent their widespread use in safety-critical applications.
no code implementations • 7 Dec 2020 • Tim Roith, Leon Bungert
In particular, we define a sequence of functionals which approximate the largest local Lipschitz constant of a graph function and prove $\Gamma$-convergence in the $L^\infty$-topology to the supremum norm of the gradient as the graph becomes denser.
no code implementations • 28 May 2020 • Leon Bungert, Martin Burger, Yury Korolev, Carola-Bibiane Schoenlieb
We study variational regularisation methods for inverse problems with imperfect forward operators whose errors can be modelled by order intervals in a partial order of a Banach lattice.
Numerical Analysis Numerical Analysis Optimization and Control 47A52, 65J20, 65J22, 65K10
1 code implementation • 1 Apr 2020 • Leon Bungert, Matthias J. Ehrhardt
Multi-modality (or multi-channel) imaging is becoming increasingly important and more widely available, e. g. hyperspectral imaging in remote sensing, spectral CT in material sciences as well as multi-contrast MRI and PET-MR in medicine.
no code implementations • 27 Feb 2019 • Leon Bungert, Martin Burger, Daniel Tenbrinck
In this work we investigate the computation of nonlinear eigenfunctions via the extinction profiles of gradient flows.
2 code implementations • 4 Oct 2017 • Leon Bungert, David A. Coomes, Matthias J. Ehrhardt, Jennifer Rasch, Rafael Reisenhofer, Carola-Bibiane Schönlieb
In this paper, we propose a method for increasing the spatial resolution of a hyperspectral image by fusing it with an image of higher spatial resolution that was obtained with a different imaging modality.