Reconstructing multiplex networks with higher-order interactions from dynamics

Abstract

Multiplex networks provide an important framework for characterizing multilevel interactions in complex systems, yet reconstructing their structures from dynamical data remains a challenging task. In this work, we propose a unified regression-based framework for multiplex network reconstruction grounded in nodal dynamics and coupling functions. By extending the node-wise regression reconstruction of Malizia et al. (2024) from single-layer systems to multiplex settings, the framework can simultaneously reconstruct intra-layer pairwise interactions, higher-order interaction structures, and one-to-one inter-layer couplings from observed time-series data. Numerical experiments based on the Hindmarsh–Rose neuronal model demonstrate that the proposed method achieves reliable reconstruction performance for directed weighted, undirected unweighted, and higher-order multiplex networks. Furthermore, applications to the Padgett–Florentine families multiplex social network confirm the robustness and generalization capability of the framework under real-world network topologies. Overall, this study provides a general and robust approach for systematically reconstructing multiplex networks and their higher-order interaction structures from time-series data.