On minimization of the empirical risk function by linear programming methods

Specifics of solving problems of minimizing the function of empirical risk are investigated in the class of linear desistion rules by linear programming methods. Conditions are found for significance of integer constraints on variables in these problems.