Identifying topology of general dynamical networks from noisy data

Abstract

The reverse engineering problem of identifying the topology of dynamical networks withtime varying intersystem coupling delays from noisy observation data is treated in thispaper. A three-stage approach is adopted to solve this identification problem: first,feedback control is applied to the network and it is driven to its equilibrium state; then, theinformation obtained from the observed steady state responses is explored and theidentification problem is converted into a mixed integer optimization problem; finally, aninference algorithm is applied to obtain all elements of the connectivity matrix.Several dynamical network examples are provided for illustration. Identificationresults show that the proposed method is robust to noise and computationallyefficient.