Structured covariance estimation. The routine takes a covariance matrix as input and returns the Toeplitz matrix that lies closest to it, in the sense that it minimizes the Kullback-Leibler divergence between the two. In