Minimizing increasing star-shaped functions based on abstract convexity (Q1304394): Difference between revisions

A real-valued function \(f\) defined on a cone \(K\subseteq \mathbb{R}^n\) is said to be \(+\infty\)-star-shaped if \(f(\lambda x)\geq \lambda f(x)\) for all \(x\in K\) and \(\lambda\in [0,1]\). The authors study ISSI (i.e., increasing \(+\infty\)-star-shaped functions defined on the nonnegative orthant from the point of view of abstract convexity. In particular, they introduce appropriate notions of subdifferentials and superdifferentials based on the fact that ISSI functions are abstract convex with respect to certain min-type functions and abstract concave with respect to certain max-type functions. These subdifferentials are used to construct an algorithm for the global minimization of ISSI functions; a convergence theorem is proved, and results of numerical experiments are reported.