Global saddle-point duality for quasi-concave programs

General and quasi-concave non-differentiable cases of the maximization of the minimun between two functions are considered. With the aid of duality theory for mathematical programming involving conjugate-like operators and by defining a bifunction we construct a new Lagrangian and generate a class of perturbations. New saddle-point theorems are presented, and equivalence is proved between the existence of a saddle-point and the existence of a certain cone-supporting property of the perturbation function. These results suggest possible improvements in multiplier methods.This work was partially supported by a grant from Control Data.