Using the \(L\)-curve for determining optimal regularization parameters (Q1347063)

It has been known for some time that a conventional use of the \(L\)-curve criterion for choice of the regularization parameter in the Tikhonov method for ill-posed linear operator equations does not lead to convergence (with respect to the error level in the data). In this paper the authors show, however, that the \(L\)-curve can be used as an alternate method of characterizing the choice of the regularization parameter by the discrepancy principle or by the order-optimal modifications of the discrepancy principle. This novel use of the \(L\)-curve therefore reproduces the known convergence results for the discrepancy-based methods.