Fundamental Dynamical Units for Physics-Informed Structural Inference from Perturbation Time-Series in Networked Systems
Not provided
Abstract
This paper introduces a framework for recovering signed interaction structures in networked dynamical systems using Fundamental Dynamical Units (FDUs) to address challenges in structural inference from perturbation time-series data.
Reality Card
The framework enables the joint recovery of interaction structure and perturbation-resolved trajectories through a physics-informed neural ordinary differential equation, validated on synthetic benchmarks.
The framework supports structural commitment and motif-prescribed intervention design, validated on synthetic benchmarks with known ground truth.
The main limitation is the reliance on synthetic benchmarks, which may not fully capture real-world complexities.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.