Tensor methods in nonlinear system modeling and identification

Abstract: Because of their intimate connection with multilinear forms, tensors arise naturally in nonlinear system representation with Volterra models. In particular, Volterra kernels with tensor structure arise in so-called block-structured models and in the construction of polynomial systems with certain approximation capabilities. As a consequence, results and concepts borrowed from multilinear algebra find application in nonlinear system modeling and identification problems. In this talk, I will present a survey of tensor-based methods which have been applied to these problems, and briefly discuss some open possibilities.