Set theoretic autoregressive spectral estimation

Abstract

Presents the set theoretic approach in autoregressive (AR) spectral estimation. Conventional AR estimates, which are based on some criterion of optimality, may violate a priori constraints on the problem. In the framework of set theoretic estimation, one produces an estimate of the regression vector which has the property of being consistent with all the available a priori knowledge. Each known property being associated with a set in the regression space, the problem is then to find a common point of these sets.<>