Nonasymptotic support recovery for high?dimensional sparse covariance matrices

We propose a general framework for nonasymptotic covariance matrix estimation making use of concentration inequality-based confidence sets. specify this the large sparse matrices through incorporation past thresholding estimators with key emphasis on support recovery. This technique goes beyond results by allowing wide range distributional assumptions merely sub-Gaussian tails. methodology can furthermore be adapted to other and settings. The usage dimension-free sets yields good theoretical performance. Through extensive simulations, it is demonstrated have superior performance when compared such methods. In context recovery, we are able false positive rate optimize maximize true recoveries.