Maximally monotone linear subspace extensions of monotone subspaces: explicit constructions and characterizations (Q353151): Difference between revisions

Let \(G\) be a set-valued function on \(\mathbb{R}^n\) and let \(\mathrm{gra}(G)\) denote the \textit{graph} of \(G\), viz., \(\mathrm{gra}(G):=\{(x,x^*) \in \mathbb{R}^n \times \mathbb{R}^n: x^* \in G(x)\}\). \(G\) is called \textit{monotone} if \(\langle x-y, x^*-y^* \rangle \geq 0\) for every \((x,x^*), (y,y^*) \in\mathrm{gra}(G)\). \(G\) is said to be \textit{maximally monotone} if \(G\) is monotone and \(G\) has no proper monotone etension in the sense of graph inclusion. \(G\) is called a \textit{linear relation} if \(\mathrm{gra}(G)\) is a subspace. For \(A,B \in \mathbb{R}^{p \times n}\), let \(V_-\) denote the eigenspace spanned by the eigenvectors corresponding to the negative eigenvalues of \(AB^T+BA^T\) and \(V_0\) denote the eigenspace spanned by the eigenvectors corresponding to the eigenvalue zero. Let \(G_1,G_2\) be linear relations defined in terms of their graphs by \[ \mathrm{gra}(G_1):=\{(x,x^*) \in \mathbb{R}^n \times \mathbb{R}^n: Ax+Bx^* =0\} \] and \[ \mathrm{gra}(G_2):=\{(B^Tu,-A^Tu): u \in V_-+V_0\}. \] The authors show that the following four statements are equivalent: (a) \(G_1\) is maximally monotone; (b) \(p=n\) and \(G_1\) is monotone; (c) \(p=n\) and \(AB^T+BA^T\) is negative semidefinite; (d) \(p=n\) and \(G_2\) is maximally monotone. This extends an earlier known result. Examples are provided as illustrations.