An exact penalty function algorithm for semi-infinite programmes (Q751775)

For semi-infinite nonlinear programming, a penalty-type method is investigated which is based on the \(L_{\infty}\)-norm. Proceeding from the global and some local maximal values of the constraint function, a quadratic approximation of the exact penalty function is constructed. By assuming that the constraint function possesses a finite number of global maximizers, the usage of the implicit function theorem can be avoided. A new iterate is obtained by solving the \(L_{\infty}\)-quadratic programming subproblem and by performing a line search. Convergence of the algorithm is shown.