Efficiently estimate log-determinants of large, sparse, positive definite matrices using sparse approximate inverses and graph splines.
