Skip to content
DraconiticResearch
Search Sources API
186,541 studies31 sourcesUpdated 23 min ago
PreprintCC BY 4.0arXiv

Stable Filters for Generative Modeling of Graph Signals

  • Martin Schmidt
  • Gonzalo Mateos
Open at sourcePDFDOICite

Abstract

Generating signals on graphs requires permutation-equivariant models that exhibit stability with respect to relative structural perturbations. While recent graph-aware Schrödinger bridge models incorporate topology information directly into their reference dynamics, it is unclear how perturbations of the graph propagate through these dynamics and affect the resulting generated distributions. In this paper, we analyze the structural stability of graph-aware continuous-time generative models whose drift combines a graph filter with a learned graph neural network. We derive explicit Wasserstein stability bounds that quantify the effect of relative graph perturbations on the generated distributions. Motivated by these bounds, we introduce a principled framework for designing stable graph filters that preserve the smoothing behavior of graph heat diffusion, while boosting structural stability. Experiments on synthetic and fMRI signals show our stable filters enhance structural robustness while matching or exceeding the generative quality of the heat equation baseline.

Related

  1. Optimal Trading Filters and the Value of Predictive Signals
    Matteo Smerlak
    arXivPreprintCC BY 4.0Full textOptimisation & controlStatistical & nonlinear physicsarXiv
  2. Stabilizing Filters for Graph Signal Generation
    Martin Schmidt, Gonzalo Mateos
    arXivPreprintCC BY 4.0Full textStatisticsarXiv
  3. From Global to Tail Convergence: Dynamics of α-Stable Score-Based Generative Models
    Ianis Hammani, Claire Boyer
    HALPreprintCC BY 4.0Full textProbabilityStatisticsHALarXiv
  4. Brauer graph algebras are closed under stable equivalence of Morita type
    Pengyun Chen, Nengqun Li, Yuming Liu +1
    arXivPreprintarXiv licenceFull textAlgebra & number theoryarXiv
  5. Generative modeling via hierarchical tensor sketching
    Yifan Peng, Yi-An Chen, Edwin Miles Stoudenmire +1
    Journal of Computational and Applied MathematicsPeer-reviewedLicence not statedFull text5 citedNumerical analysis & scientific computingOpenAlex

Cite

BibTeXRIS
BibTeX
@misc{schmidt2026stable,
  title = {{Stable Filters for Generative Modeling of Graph Signals}},
  author = {Martin Schmidt and Gonzalo Mateos},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.18759},
  eprint = {2609.18759},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.18759},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - Stable Filters for Generative Modeling of Graph Signals
AU  - Martin Schmidt
AU  - Gonzalo Mateos
PY  - 2026
DA  - 2026/09/16
PB  - arXiv
DO  - 10.48550/arxiv.2609.18759
UR  - https://arxiv.org/abs/2609.18759
AB  - Generating signals on graphs requires permutation-equivariant models that exhibit stability with respect to relative structural perturbations. While recent graph-aware Schrödinger bridge models incorporate topology information directly into their reference dynamics, it is unclear how perturbations of the graph propagate through these dynamics and affect the resulting generated distributions. In this paper, we analyze the structural stability of graph-aware continuous-time generative models whose drift combines a graph filter with a learned graph neural network. We derive explicit Wasserstein stability bounds that quantify the effect of relative graph perturbations on the generated distributions. Motivated by these bounds, we introduce a principled framework for designing stable graph filters that preserve the smoothing behavior of graph heat diffusion, while boosting structural stability. Experiments on synthetic and fMRI signals show our stable filters enhance structural robustness while matching or exceeding the generative quality of the heat equation baseline.
ER  -