Stable iterative Lagrange principle in convex programming as a tool for solving unstable problems (Q2357119): Difference between revisions

The authors consider a special class of convex optimization problems in Hilbert spaces. The problems involve a strongly convex goal function, both equality and inequality constraints with parameters and admit perturbations of all the functions. The authors analyse the dual function and its sub-differential defining the corresponding Lagrange multipliers. They consider the usual and iterative regularization methods for the perturbed dual problem and prove different convergence properties depending on solvability of the non-perturbed dual problem. Besides, certain concordance rules for perturbations and regularization parameters are necessary. The approach is illustrated by application to a Fredholm integral equation.