1: \begin{abstract}

2: The theory of network identification, namely identifying the (weighted) interaction topology among a known number of

3: agents, has been widely developed for linear agents. However, the theory for nonlinear agents using probing inputs is less developed orange and relies on dynamics linearization. We use global convergence properties of the network, which can be assured using passivity theory, to present a network identification method for nonlinear agents. We do so by linearizing the steady-state equations rather than the dynamics, achieving a sub-cubic time algorithm for network reconstruction for nonlinear agents and controllers. We also study the problem of network reconstruction from a complexity theory standpoint, showing that the presented algorithms are optimal in terms of time complexity. We provide examples of reconstructing large-scale networks, including a network of first-order linear agents, and a non-linear neural network model.

4: \end{abstract}

5: