Detecting structural perturbations from time series with deep learning

Deep learning of contagion dynamics on complex networks

“…More specifically, deep learning is now being applied to problems related to Network Science using tools like Graph Neural Networks (GNN) which have been designed to handle arbitrarily structured data [20][21][22]. Recent work showed great promise for applications in community detection and link prediction [23,24], in the prediction of dynamical observables [25], in network reconstruction [26] and the detection of structural perturbations [27], as well as in the context of discovering new materials and drugs [28,29]. In general, GNNs have been shown to adequately handle in-tricate tasks, making them prime candidates to tackle several challenges of contagion dynamics modeling.…”