Lexicographic regularization and duality for improper linear programming problems (Q2396379)

Improper problems are the problems of linear and convex programming for which the basic relations of duality are violated; namely, the simultaneous feasibility of the primal and dual problems and the coincidence of their optimal values do not hold. An improper problem has no solution in the usual meaning of the term. Therefore, it must be subjected to some correction or, in other words, to a structural or informational transformation that reduces this problem to a feasible form. The authors propose a new approach to the lexicographic correction of improper linear programming problems, based on the multistep symmetric regularization of the classical Lagrange function with respect to primal and dual variables simultaneously. The convergence conditions for the method are presented, and an informal interpretation of the obtained solution is given.