我们在之前用压缩感知方法探测隐藏节点的工作基础上，继续发展了这套方法。主要思路是区分噪音和隐藏节点的区别，并且可以探测多个隐藏节点。 http://www.nature.com/srep/2014/140203/srep03944/full/srep03944.html Ascertaining the existence of hidden objects in a complex system, objects that cannot be observed from the external world, not only is curiosity-driven but also has significant practical applications. Generally, uncovering a hidden node in a complex network requires successful identification of its neighboring nodes, but a challenge is to differentiate its effects from those of noise. We develop a completely data-driven, compressive-sensing based method to address this issue by utilizing complex weighted networks with continuous-time oscillatory or discrete-time evolutionary-game dynamics. For any node, compressive sensing enables accurate reconstruction of the dynamical equations and coupling functions, provided that time series from this node and all its neighbors are available. For a neighboring node of the hidden node, this condition cannot be met, resulting in abnormally large prediction errors that, counterintuitively, can be used to infer the existence of the hidden node. Based on the principle of differential signal, we demonstrate that, when strong noise is present, insofar as at least two neighboring nodes of the hidden node are subject to weak background noise only, unequivocal identification of the hidden node can be achieved. srep03944.pdf
标签: 网络重构
