Global optimality conditions and exact penalization (Q2421449)

The author considers nonconvex optimization problems with inequality constraints. Both the objective function and the functions occuring in the constraints are d.-c. functions, i.e., they are expressed as the difference of two convex functions. All functions of the problems are differentiable. It is further assumed that the set of feasible solutions is non-empty, the variables of the problems belong to a closed convex set and the set of global solutions is non-empty. Under these conditions, the original problem is reduced to a problem without inequality constraints by making use of the exact penalization procedure. The objective function of the penalized problem has the d.-c. form. Necessary and sufficient global optimality conditions are proved. Theoretical results are demonstrated on small numerical examples.