Topology Identification of Weighted Networks Via Binary Time Series From Propagation Dynamics

This study focuses on a topology identification problem of weighted networks with different connection strength, where binary time series generated by propagation dynamics are utilized. An influence probability matrix reflecting the weight of connection is proposed to quantify the influence of other nodes on one node as it transfers from susceptible state to infected state. Further, maximum likelihood estimate and expectation–maximization algorithm are used to obtain the influence probability matrix. A threshold method and a weight-based-identification algorithm are provided to identify connection strength. The robustness against fault data and conflicting results of the same connection is mitigated by introducing a confidence factor. Several Monte-Carlo simulations demonstrate the high identification accuracy of our methods under different network models.