Linear equations and inequalities on finite dimensional, real or complex, vector spaces: a unified theory (Q2532626): Difference between revisions

This paper presents a unified theory of linear equations and inequalities, real or complex, on finite-dimensional vector spaces. Notations: Let \(S\subset \mathbb C^n\) be a closed convex cone, \[ S^*= \{y\in\mathbb C^n : x\in S\Rightarrow \text{Re}(y^Hx)\ge 0\},\ A\in \mathbb C^{m\times m}, b\in \mathbb C^m. \] The system \((*)\) \(Ax = b\), \(x\in S\) is ``consistent'' [ ``asymptotically consistent'' ] if \(\exists\, x \in S \ni Ax = b\) [ \(\exists\, \{x_k\}\subset S \ni \lim Ax_k = b\) ]. Sample results: Theorem: The following are equivalent (where \(A^T, A^H, A^+\) denote the transpose, the conjugate transpose, the generalized inverse of \(A\)):: (a) \((*)\) is asymptotically consistent, (b) \(A^H y\in S^* \Rightarrow \text{Re}(b^H y)\ge 0\), (c) \(b\in R(A)\) and \(A^+ b\in \text{cl}(N(A) + S)\). Corollary: Let \(N(A)+S\) be closed. Then \((*)\) is consistent if, and only if, \(A^H y\in S^* \Rightarrow \text{Re}(b^H y) = 0\). Corollary (Solvability of linear equations): \(Ax = b\) is consistent if, and only if, \(A^H y = 0 \Rightarrow b^H y = 0\). Corollary (Farkas): Let \(A\in \mathbb R^{n\times n}\), \(b\in\mathbb R^n\). Then \(Ax =b\), \(x \ge 0\) is consistent if, and only if, \(A^T y\ge 0 \Rightarrow b^T y\ge 0\). These results are used to prove a duality theorem for complex linear programming, following and slightly extending \textit{N. Levinson} [J. Math Anal. Appl. 14, 44--62 (1966; Zbl 0136.13802)].