On improving accuracy for Arcangeli's method for solving ill-posed equations (Q5937213)

The authors treat Tikhonov regularization for solution of the ill-posed linear operator equation \(Tx= y\), where \(T\) is a bounded linear operator between Hilbert spaces having a non-closed range. For given data \(y^\delta\) satisfying \(\|y- y^\delta\|\leq\delta\), and assuming that the minimal norm least squares solution is in the range of the operator \((T^*T)^\nu\) for some \(\nu\in (0,1/2)\), the authors show that a slightly modified form of Arcangeli's discrepancy principle for the choice of the regularization parameter yields an asymptotic convergence rate of \(O(\delta^{\nu/(\nu+ 1)})\). This improves the best previously known rate of Arcangeli's method in this case.