Functional convergence theorems for the penalty function method (Q1964470)

In problems of convex programming it is very important to obtain sufficient conditions of convergence of the processes while using the penalty function method. Two sufficient conditions for convergence of the penalty function method in relation to a functional are presented. It is shown that two conditions, i.e. the convergence of the penalty function method and the semicontinuity of the perturbation function of the initial problem are equivalent. Classes of problems for which optimal sets of proper approximating problems do not tend to infinity from the origin of coordinates while a penalty coefficient tends to infinity are found.