An approach to parameter identification for Boolean-structured multilinear time-invariant models*

In this paper, a parameter identification method for multilinear time-invariant grey-box models which are structured by Boolean functions is presented. Applying binary indexing to network graphs enables an efficient extraction of structural system information which can be transformed into continuous Zhegalkin polynomials. Their parameters are identified with a nonnegative alternating least squares (ALS)-algorithm. A running example demonstrates the application of this method and the algorithm is briefly analysed.