arXiv subjects

Christian Koke

Publications and source records attributed to Christian Koke.

8 recordsLinked to original sources

arXiv2026

Graph Neural Networks Are Not Continuous Across Graph Resolutions

We show that contrary to conventional wisdom in the community, graph neural networks (GNNs) are not continuous with respect to all natural modes of graph convergence. As a result, GNNs may generate substantially different latent representations for graphs that are very similar. In particular they assign vastly different latent embeddings to graphs that represent the same underlying object at different resolution scales. We trace this failure of continuity back to a structural obstruction arising from commonly used information-propagation schemes. Building on this insight we then derive a principled modification to standard GNN architectures which equips models with continuity across scales. The proposed modification enables consistent integration of distinct resolutions and reliable generalization between them. We systematically validate our theoretical findings in a wide range of numerical experiments.

cs.LG↗

arXiv2026

Large Coupling Convergence Beyond Definiteness

We study convergence of operator families of the form $A_\beta = A + \beta B$ towards an effective operator defined on $\ker(B)$, as the coupling constant $\beta$ tends to infinity. Crucially, we focus on the setting where neither $A$ nor $B$ can be assumed to be positive- (or negative-) semi-definite. We are hence outside the classical form-theoretic framework, where results based on Kato's monotone convergence theorem would be applicable. Thus, instead of form methods, our approach builds on classical resolvent identities to study convergence of the family $\{A_\beta\}_{\beta}$. Our findings are that: (i) \emph{Strong} resolvent convergence holds (without further spectral assumptions) if $A + \beta B$ is self-adjoint and the compression of $A$ onto $\ker(B)$ is well behaved. (ii) Under the more detailed assumption that $0 \in \sigma(B)$ is isolated, \emph{norm} resolvent convergence can be established even if $A+\beta B$ is merely closed, provided the quasinilpotent part of $B$ at zero vanishes and certain conditions on the interplay of $A$ and $B$ are met. Importantly, if $B$ is not self-adjoint we find that the limit operator not only depends on $\ker(B)$ as a Hilbert space, but crucially also on the precise form of the Riesz projector at $0 \in \sigma(B)$ onto $\ker(B)$.

math.FA↗

arXiv2026

Di-Graphs with tightly connected Clusters: Effective Graph Laplacians and Resolvent Convergence

In this note, we study Laplacians on graphs for which connectivity within certain subgraphs tends to infinity. Our main focus are graphs sharing a common node set on which edge weights within certain clusters grow to infinity. As intra-cluster connectivity increases, we show that the corresponding graph Laplacians converge -- in the resolvent sense -- to an effective graph Laplacian. This effective limit Laplacian is defined on a coarsened graph, where each highly connected cluster is collapsed into a single node. In the undirected setting, the effective Laplacian arises naturally from aggregating over tightly connected clusters. In the directed case, the limiting graph structure depends on the precise manner in which connectivity increases; with the corresponding effects mediated by the left and right kernel structure of the Laplacian restricted to high-connectivity clusters. Our results shed light on the emergence of coarse-grained dynamics in large-scale networks and contribute to spectral graph theory of directed graphs.

math.FA↗

arXiv2023

HoloNets: Spectral Convolutions do extend to Directed Graphs

Within the graph learning community, conventional wisdom dictates that spectral convolutional networks may only be deployed on undirected graphs: Only there could the existence of a well-defined graph Fourier transform be guaranteed, so that information may be translated between spatial- and spectral domains. Here we show this traditional reliance on the graph Fourier transform to be superfluous and -- making use of certain advanced tools from complex analysis and spectral theory -- extend spectral convolutions to directed graphs. We provide a frequency-response interpretation of newly developed filters, investigate the influence of the basis used to express filters and discuss the interplay with characteristic operators on which networks are based. In order to thoroughly test the developed theory, we conduct experiments in real world settings, showcasing that directed spectral convolutional networks provide new state of the art results for heterophilic node classification on many datasets and -- as opposed to baselines -- may be rendered stable to resolution-scale varying topological perturbations.

cs.LG↗

arXiv2023

ResolvNet: A Graph Convolutional Network with multi-scale Consistency

It is by now a well known fact in the graph learning community that the presence of bottlenecks severely limits the ability of graph neural networks to propagate information over long distances. What so far has not been appreciated is that, counter-intuitively, also the presence of strongly connected sub-graphs may severely restrict information flow in common architectures. Motivated by this observation, we introduce the concept of multi-scale consistency. At the node level this concept refers to the retention of a connected propagation graph even if connectivity varies over a given graph. At the graph-level, multi-scale consistency refers to the fact that distinct graphs describing the same object at different resolutions should be assigned similar feature vectors. As we show, both properties are not satisfied by poular graph neural network architectures. To remedy these shortcomings, we introduce ResolvNet, a flexible graph neural network based on the mathematical concept of resolvents. We rigorously establish its multi-scale consistency theoretically and verify it in extensive experiments on real world data: Here networks based on this ResolvNet architecture prove expressive; out-performing baselines significantly on many tasks; in- and outside the multi-scale setting.

cs.LG↗

arXiv2023

Limitless stability for Graph Convolutional Networks

This work establishes rigorous, novel and widely applicable stability guarantees and transferability bounds for graph convolutional networks -- without reference to any underlying limit object or statistical distribution. Crucially, utilized graph-shift operators (GSOs) are not necessarily assumed to be normal, allowing for the treatment of networks on both undirected- and for the first time also directed graphs. Stability to node-level perturbations is related to an 'adequate (spectral) covering' property of the filters in each layer. Stability to edge-level perturbations is related to Lipschitz constants and newly introduced semi-norms of filters. Results on stability to topological perturbations are obtained through recently developed mathematical-physics based tools. As an important and novel example, it is showcased that graph convolutional networks are stable under graph-coarse-graining procedures (replacing strongly-connected sub-graphs by single nodes) precisely if the GSO is the graph Laplacian and filters are regular at infinity. These new theoretical results are supported by corresponding numerical investigations.

cs.LG↗

arXiv2023

Graph Scattering beyond Wavelet Shackles

This work develops a flexible and mathematically sound framework for the design and analysis of graph scattering networks with variable branching ratios and generic functional calculus filters. Spectrally-agnostic stability guarantees for node- and graph-level perturbations are derived; the vertex-set non-preserving case is treated by utilizing recently developed mathematical-physics based tools. Energy propagation through the network layers is investigated and related to truncation stability. New methods of graph-level feature aggregation are introduced and stability of the resulting composite scattering architectures is established. Finally, scattering transforms are extended to edge- and higher order tensorial input. Theoretical results are complemented by numerical investigations: Suitably chosen cattering networks conforming to the developed theory perform better than traditional graph-wavelet based scattering approaches in social network graph classification tasks and significantly outperform other graph-based learning approaches to regression of quantum-chemical energies on QM7.

cs.LG↗

arXiv2020

Dirac equation on a square waveguide lattice with site-dependent coupling strengths and the gravitational Aharonov-Bohm effect

The main objective of this work is to present a theoretical proposal for an implementation of the $(2 + 1)$-dimensional Dirac equation in classical gravitational and electromagnetic backgrounds in a two-dimensional waveguide array. For this, a framework for achieving site-dependent effective coupling constants in two-dimensional waveguide arrays is developed. Implementability of the Dirac equation under the proposed scheme puts minor demands on gauge and spacetime backgrounds; however, a wide array of physical spacetimes, such as all vacuum and static solutions, prove to be implementable. As an interesting and instructive example, we discuss a tabletop realization of the gravitational Aharonov-Bohm effect: After devising a thought experiment in which signatures of the gravitational Aharonov-Bohm effect could be detected, we briefly discuss how the analogue of such a setting can in principle be implemented using the proposed waveguide setup.

quant-ph↗