Order generalised gradient and operator inequalities (Q2261988)

A vector \(x^*\) is said to be a subgradient of a convex function \(f : K \subseteq \mathbb{R}^n \to \mathbb{R}\) at a point \(x\) if \[ f (x) - f (y) \geq x^* (x - y) \tag{\(\diamondsuit\)} \] for all \(y \in K\). Inequality (\(\diamondsuit\)) is called the subgradient inequality and its validity for all \(x \in K\) characterizes the convexity of \(f\). The authors introduce the notion of order generalized gradient for operator-valued functions, which is a generalization of the subgradient, without the assumption of convexity, as follows: Let \(\mathcal{C}\) be a convex set in the real space \(B(H)_h\) of all self-adjoint bounded linear operators acting on a Hilbert space \(H\) equipped with the usual order. A function \(f: \mathcal{C} \to B(H)_h\) has the function \(\nabla_f: \mathcal{C} \times B(H)_h \to B(H)_h\) as an order generalized gradient if \(f(A)-f(B) \geq \nabla_f(B, A-B)\) for any \(A, B \in \mathcal{C}\). Furthermore, the authors state some inequalities of Hermite-Hadamard, Jensen and Slater types for the order generalized gradient. They also state the connection between the order generalized gradient and Gâteaux derivative of operator-valued functions and give a characterization of convexity analogues to (\(\diamondsuit\)) in the context of operator-valued functions.