Building on this, the estimator introduces a nonconvex regularization-based sparse SCM by constraining the coherence matrix to have potential sparsity and low-rank (LR), which suppresses the influence ...