Toeplitz Covariance Matrices and the von Neumann Relative Entropy
Tryphon T. Georgiou · Birkhäuser Boston eBooks · 2003
Sample covariances based on time-series data fail to be Toeplitz. The purpose of the paper is to suggest the von Neumann relative-entropy as a distance measure for approximating a positive-definite sample covariance by one having the Toeplitz structure. This leads to a convex optimization problem whose solution retains the property of being positive-definite.