“…It is well-known that the entries of matrix functions f (A) often exhibit an exponential or even super-exponential decay away from the sparsity pattern of A: The larger the geodesic distance d(i, j) of node i and j in the graph of A, the smaller the entry [f (A)] ij can be expected to be. This was first studied for the inverse of banded matrices in [27] and later extended to other functions and matrices with more general sparsity pattern in numerous works; see, e.g., [14,9,15,16,32,33,46,48,49] and the references therein; Specifically, in [47] such an approach was applied in the context of network modifications.…”