Identifiability and convergence analysis of the MINLIP estimator

Convex Optimization for Blind Identification of Monotone Wiener Systems**This work was supported by the Swedish Research Council under contract 621-2007-6364.

“…We note that the convergence analysis as presented in Dai et al (2015) is still valid, since this one only uses the geometry of the constraints. These are exactly the same in MINLIP * as in MINLIP.…”

“…Note that those functions are often not invertible, so that the 'inverse approach' of rephrasing the Wiener ID task as (simpler) Hammerstein ID problem is not feasable. Practical issues Ahmed et al (2013) as well as theoretical properties of MINLIP are investigated and published on in some detail, see Dai et al (2015). Formally, the model becomes…”

“…Identification (ID) of such systems aims at identifying both blocks from observation of inputoutput behaviour. The fact that this is uniquely possible under ideal circumstances, is treated in the research on identifiability as studied in Dai et al (2015), going back to work in the statistical community on additive index models Lin and Kulasekera (2007). Much research work has been devoted to designing and testing algorithms to perform such identification in a efficient, noise-tolerant, analytical tractable way, see e.g.…”

“…A simple recursive version was studied in Pelckmans and Dai (2011). Theoretical issues as identifiability and persistency of excitation were studied in Dai et al (2015). Application of such systems towards control was studied in Dai and Pelckmans (2012).…”