ON THE FINITE EPSILON-CONVERGENCE OF EXTERIOR PENALTY METHODS

K. Truemper [1975] proved that it is possible to select a target value mu > 0 for the penalty parameter of the exterior function P(r)(x,mu) = f(x) + muSIGMA(i=1)m max(0, -g(i)(x))r where r = 2 such that when m u is increased to this value or becomes larger, then x which minimizes P(r)(x,mu) is within a specified tolerance from the optimal solution of the problem min f(x) subject to g(i)(x) greater-than-or-equal-to 0 , FOR-ALLi = 1,...,m. In this paper, we extend his results for r > 1.