Application of theorems on alternatives to the determination of normal solutions of linear systems (Q1405714)

In this article the problem of solving linear systems of equalities and inequalities is studied. An approach for solving such systems based on a constructive proof of theorems on alternatives is suggested. The main mathematical tool used in this approach is the duality theory for nonlinear programming. Given an initial system of linear inequalities and equalities an alternative system is introduced such that the solvability problem is reduced to the unconstrained minimization of the norm of the discrepancy of either the initial or the alternative system. As an example for the use of the suggested approach the authors consider optimality conditions (different from the classical Karush-Kuhn-Tucker ones) for linear programming. After a single unconstrained minimization of a smooth pieceweise quadratic function the normal solution of the linear programming and the optimal discrepancy with the minimal orm of the dual problem are obtained.