Dresden 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

DY: Fachverband Dynamik und Statistische Physik

DY 31: Networks, From Topology to Dynamics – Part I (joint session SOE/DY)

DY 31.3: Talk

Wednesday, March 11, 2026, 10:30–10:45, GÖR/0226

Learning collective variables for time-evolving networks — •Sören Nagel, Nataša Djurdjevac Conrad, Stefanie Winkelmann, and Marvin Lücke — Zuse Institute Berlin

We address the challenge of model reduction for time-evolving networks by identifying collective variables for stochastic rewiring processes driven by opinion homophily. [Lücke et al., Phys. Rev. E 109, L022301 (2024); Djurdjevac Conrad et al., Chaos 34, 093116 (2024)].

Utilizing the transition manifold framework, we identify a simple consensus measure as a collective variable for an ergodic and a non-ergodic model, and learn the dynamics of the projected system. We show that the learned model reduction can be obtained from the corresponding graphon process in the case of large and not too sparse graphs with uniformly distributed opinions. Our data-driven approach successfully identifies the collective variables in more general cases, highlighting the possibility to study low-dimensional model reductions in systems that have not been understood theoretically.

Keywords: Model Reduction; Temporal Graphs; Opinion Dynamics