OPTIMIZATION OF MIN-MAX PROBLEMS WITH NON-DIFFERENTIABLE FUNCTIONS

Abstract

The paper considers a minimization problem where the objective is the sup-norm of a linear mapping with linear constraints. The methods presented are based on the con- cepts and operations of the adaptive method of linear pro- gramming with a new variant support matrix called coor- dinator support but at the same time specific features of Min-Max problems will be taken into consideration. Here the optimality and the suboptimality criteria are formulated. The iteration from an interior point allows to obtain an ε-optimal solution with a precision ε ≥ 0 chosen in advance. Results of numerical experiments are presented which include a pro- gram of this approach.