Publications
You can also find my articles on my Google Scholar profile.
Towards an Optimal Control Foundation of Early Exits – Turnpikes and ResNets
Published in 2026 The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026
In this work we leverage the toolset from control theory to give a theoretical grounding of the joint training formulation of early exit based ResNets. We introduce the notion of a Turnpike phenomena in the infinite depth regime of ResNets, which naturally translates to the empricial observations of early-exit trained networks supporting the success of intermediate classifiers and thresholding selection. Download paper here
Port-Hamiltonian Neural Networks for Systems with Multiple Asymptotically Stable Equilibria
Published in 2026 NeurIPS Workshop: AXIOM, 2026
Here we propose a stable neural network that is not restricted to be globally asymptotically stable with respect to one equilibrium. We give a constructive formulation of a port-Hamiltonian neural network that is almost-everwhere stable with respect to a discrete set of equilibria. Download paper here
Get a GRIP, this will be a long TRIP: A Quantifiable Long-Range Framework for Verifying Over-squashing
Published in 2026 The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026
In this work we propose a minimal set of axioms that a benchmark should obey when testing long-range interactions on graphs. Building upon statistical approximation, strictly ranged problems should be available which can be measured by mutual information within the receptive field of a GNN. Download paper here
Port-Hamiltonian Architectural Bias for Long-Range Propagation in Deep Graph Networks
Published in 2025 International Conference on Learning Representations (ICLR), 2025
This work tackles the challenge of long-range information propagation in graph neural networks. We proposed port-Hamiltonian Deep Graph Networks, a novel framework inspired by physics that models neural information flow using principles from Hamiltonian systems. By combining conservative and dissipative dynamics, we control how information spreads or fades across the graph. Despite its theoretical grounding, the method integrates easily into standard message-passing networks and shows strong empirical performance, especially in tasks that require effective long-range reasoning. Download paper here
Memory Efficient Kernel Approximation for Non-Stationary and Indefinite Kernels
Published in 2022 International Joint Conference on Neural Networks (IJCNN), 2022
This paper originated from my bachelor thesis and is all about large scale kernel methods. We explored the memory efficient approximation of kernel matrices that lead to invalid kernels. We proposed to correct the eigenspectrum with a shift computed via the Lanczos-Iteration. This correction leads to stable downstream tasks and despite having a large approximation error, competitive scores on the tasks. Download paper here
