A Pathwise Algorithm for Covariance Selection

Abstract

Covariance selection seeks to estimate a covariance matrix by maximum likelihood while restricting the number of nonzero inverse covariance matrix coefficients. A single penalty parameter usually controls the tradeoff between log likelihood and sparsity in the inverse matrix. We describe an efficient algorithm for computing a full regularization path of solutions to this problem.

More details & numerical experiments. Not all figures could be uploaded on arXiv. Please get local pdf file for complete numerical results