A geometrical insight on pseudoconvexity and pseudomonotonicity (Q849330)

The paper introduces the notion of \(g\)-pseudoconvex functions (where \(g\) stands for geometric): a function \(f:\mathbb R^n\to [-\infty,\infty]\) is \(g\)-pseudoconvex if it is quasiconvex and for each \(x\in\mathbb R^n\) with \(f(x)>\inf f\), one has \(\emptyset\neq \operatorname{int}S(x)\subseteq \widetilde{S}(x)\) and \(S(x)\subseteq \operatorname{cl}\widetilde{S}(x)\); here, \(S(x)= \{y\in \mathbb R^n:f(y)\leq f(x)\}\) and \(\widetilde{S}(x)=\{y\in \mathbb R^n:f(y)<f(x)\}\) are the level set and the strict level set corresponding to \(x\), respectively. These functions are studied and their relation to the properties of the ``normal cone map'', especially maximal pseudomonotonicity, is established. The normal cone map is the set-valued map \(N:\mathbb R^n\to 2^{\mathbb R^n}\), where \(N(x)\) is the normal cone to \(S(x)\) at \(x\).