Dual convergence for penalty algorithms in convex programming (Q430935)

The authors consider the class of constrained convex optimization problems with smooth functions under the Slater constraint qualification condition. They first investigate the property of convergence of multipliers in a general class of penalty methods under certain primal convergence properties. The new sufficient condition seems weak, but need additional efforts for the verification. Further a similar convergence property is established for a class of combined proximal penalty methods. Some illustrative computational results are also given.