Lucas Morisset

We introduce a data-centric estimator of the squared error of usual precision matrix estimates, both in the non-augmented case and the augmented case. Furthermore, we give non-asymptotic guarantees on our estimate.