Choice of parameters in the method of regularization of \(L\)-pseudoinversion (Q1405710)

The authors investigate a regularization method for linearly constrained least squares problems which may be loosely described as finding, among all least squares solutions of given linear operator equation of the first kind involving a closed linear operator, a minimal norm least squares solution of another linear operator equation of the same kind. It is assumed that the two operators satisfy the so-called generalized complementarity condition -- essentially that the Cartesian product operator is bounded below. A regularized approximation to the solution of the optimization problem is obtained as a minimizer of a Tikhonov-like functional involving two parameters. A pair of discrepancy-like conditions is used to successively chose the two regularization parameters and the stability and convergence of the resulting method is proved.