The Impact of Network Structure on Social Contagion and Neural ODE Generalization

Event: Max Plank Institute for Intelligent Systems: Social Foundations of Computation, Tübingen, DE

Date: 2026/10/06

Many thanks to Jiduan Wu for inviting me and to everyone in the research group for the inspiring conversations.

Abstract

A system’s structure, often represented as a network, shapes not only how dynamical processes unfold but also how well we can learn and predict them. First, I will explore the influence of structure on dynamics in the context of systems that exhibit both homophily, i.e., a preference of individuals that are similar to interact, and group interactions, represented as hyperedges that connect two or more individuals. A mathematical model of information spreading in such systems reveals that how homophilic preferences change with interaction size determines inequalities in information access between individuals with different attributes. Second, I will turn to the impact of network structure on generalization of neural ODEs. Using a random graph model to create a controlled experimental setting, I show that degree heterogeneity is a key factor determining how well neural ODEs generalize across networks of different sizes and structures, while clustering plays a secondary role.

Slides

Citation

When using these materials, please cite:

@article{laberdiesehlert2025_effectshigherorderinteractions,
  title = {Effects of Higher-Order Interactions and Homophily on Information Access Inequality},
  author = {Laber, Moritz and Dies, Samantha and Ehlert, Joseph and Klein, Brennan and Eliassi-Rad, Tina},
  year = {2025},
  journal = {Communications Physics},
  volume = {9},
  number = {1},
  pages = {13},
  doi = {10.1038/s42005-025-02445-y}
}

@misc{laber2026_guidehigherorderhomophily,
  title = {A Guide to Higher-Order Homophily},
  author = {Laber, Moritz and Klein, Brennan},
  year = 2026,
  archiveprefix = {arXiv}
  eprint = {2606.02537},
  doi = {10.48550/arXiv.2606.02537}
}

@misc{laber2026_neuralODE,
  title = {When do neural ordinary differential equations generalize on complex networks},
  author = {Laber, Moritz and Klein, Brennan and Eliassi-Rad, Tina},
  year = {2026},
  archiveprefix = {arXiv}
  eprint = {2602.08980},
  doi = {10.48550/arXiv.2602.08980}
}