scientific article; zbMATH DE number 3898622

From MaRDI portal
Publication:3677521





[For the entire collection see Zbl 0538.00033.] We consider the lexicographical order relation for vectors and matrices in view of its applications to optimization problems. A nonnecessarily square matrix is said to be lexicographically nonnegative if each of its columns is lexicographically non-negative. This is proved to be equivalent to the existence of a decomposition of A into a product of a unitary lower triangular matrix and a termwise nonnegative matrix. From this result we obtain a Farkas type theorem for lexicographical consequences of linear inequality systems. In the case of convex systems we give a characterization of inconsistency which does not require the lower semicontinuity of the functions. We give three applications to optimization: the proof of the existence of a matricial Lagrange multiplier of a suitably defined vector Lagrangian for an optimal solution of a convex problem (with no constraint qualification), a characterization of the Chebyshev solutions of an inconsistent problem and an improvement of the formulation of the exact quasiconvex duality theory [see the author, Z. Oper. Res., Ser. A 27, 257-266 (1983; Zbl 0522.90069)]. A further improvement of this last theory has been obtained in another recent paper by the author [Method. Oper. Res. 49, 47-62 (1985)].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3677521)