Model identification of a network as compressing sensing

In many applications, it is of interest to derive information about the topology and the internal connections of multiple dynamical systems interacting together. Examples can be found in fields as diverse as Economics, Neuroscience and Biochemistry. The paper deals with the problem of deriving a descriptive model of a network with no a-priori knowledge on its topology. It is assumed that the network nodes are passively observed and data are collected in the form of time series. The underlying structure is then determined by the non-zero entries of a "sparse Wiener filter". We cast the problem as the optimization of a quadratic cost function, where a set of parameters are used to operate a trade-off between accuracy and complexity in the final model.