摘要：AbstractThis paper studies a Convex Optimization (CO) method to learn a monotone Wiener system, consisting of (i) a linear state-space model of finite order corresponding to the LTI block of the system, and of (ii) a monotone output nonlinearity. This work advances the previously published work of the author on the MINLIP method towards state-space descriptions. The motivation to study Wiener systems is that they are arguably the simplest, non-trivial examples of nonlinear systems, and that they provide a forum to develop techniques which can be used for identifying Volterra, LFTs or other more advanced nonlinear dynamical models. The key idea is to use the nuclear norm of the Hankel matrix corresponding to the model used for describing the LTI block. The approach is extended towards the use of CO for the blind identification of such systems.
