Inverse Network Dynamics - Network structure and function from nonlinear dynamics and time series data

NetDat22 took place at the MPIPKS from September 12-21, 2022. The aim of the event was to bring together theory and method development on inverse problems in network dynamics, on the rapidly evolving topics of network structural inference and on the network design for desired dynamics, as well as their numerous fields of application. The workshop had almost 100 participants from 19 countries working in the areas of Nonlinear Network Dynamics, Time Series Analysis, Statistical Inference as well as Machine Learning and Data-driven Modeling.
Given the event has been rescheduled and newly planned twice from 2020 to 2021 to 2022, we adapted the original schedule of starting with 2.5 days seminar lectures followed by a week of workshop about new research results. To accommodate time constraints by the speakers (in part online, in part on-site), the event offered seminar sections distributed across the entire 1.5 weeks. Here, several established researchers presented introductory material about the full range of topics: higher-order interactions in multi-dimensional dynamical systems, universal aspects of network dynamics, inference of oscillator responses from probing experiments, approaches and challenges for the inference of various functional, effective or structural network from data, and a guided tour to (deep) machine learning for non-machine learning researchers, among others.

Many eminent contributors, along with young scientists, presented their topical research findings in the workshop parts. Common topics included the dynamics of (and near) tipping points and their prediction, avoidance as well as recovery of normal or desired operating states; the conservation of Granger causality by suitable compressed sensing matrices, interpretable machine learning for learning network structure; the inference of network size and structure based on multiple transient time series; the mathematical option space for reconstructing network topologies from the time series of their node dynamics by sparse regression; and the quantification by a sensitivity measure for inference problems by bridging exact mathematical work with physics-style exploration for cases not tractable analytically. Posters were available also after a poster session catalyzing several additional discussions.

NetDat22 had a big contribution towards community building, as indicated by many lively discussions not only at the dedicated discussion sessions, but also during the coffee breaks, lunch or dinner as well as a discussion dinner self-organized by the participants on the final evening of the event. The event took place in a hybrid format, yet the workshop succeeded in giving a strong impulse for collaborative research. The participants appreciated the topical breadth of the program. Many excellent presentations of the different approaches developed across wide-spread application fields generated an extraordinary number of interactions between participants from different fields and backgrounds, initializing new international and interdisciplinary collaborations.

We would like to thank the MPIPKS for providing the opportunity for this highly relevant cross-disciplinary exchange, Katrin Lantsch for providing outstanding organizational support and Ronny Börner for using NetDat22 insights and more on all IT tasks during the workshop. Thank You!