Estimation of the multivariate distributions precision matrix learning

In this paper, we give a estimation of the precision matrix learning of multivariate Gaussian distributions. The conclusion is that we can turn the approximation of the precision matrix to the approximation of the covariance matrix, and prove the asymptotic converges of the algorithm using σ_k=0ⁿ(I-A_z)^k to approximate the precision matrix A^-1.