Topology identification of a sparse dynamic network

This paper addresses the problem of identifying the topology of a sparsely connected network of dynamic systems. The goal is to identify the links, the direction of information flow and the transfer function of each dynamic system. The output of each system is affected by the incoming data of the directly connected systems and noise. In contrast to the related existing work we use causal Laguerre basis functions to expand the transfer functions. Since the network is sparsely connected we estimate the system topology using an algorithm which optimizes the l₀ penalized least squares criterion with grouped variables. This also contrasts with previous work which usually uses and l₁ penalty. The l₀ penalty has the potential to generate greater sparsity. We present simulation results to demonstrate the effectiveness of this method.