A new exclusion test for finding the global minimum (Q869487)

The following minimization problem is considered: minimize \(f(x)\) subject to \(x\in\Lambda\), where \(f:\mathbb R^n\to \mathbb R\) is a Lipschitz function of order \(\alpha\), \(\alpha\geq 1\), and \(\Lambda\) is an \(n\)-dimensional rectangular box (a cell). A new exclusion algorithm for finding a global minimum of function \(f\) on set \(\Lambda\) is proposed. Exclusion algorithms are algorithms, which systematically discard cells, in which a necessary condition for the presence of a global minimum is not satisfied. This necessary condition is called minimization condition. Different exclusion algorithms make use of different minimization conditions. The minimization conditions known from the literature either involve derivatives or can be applied to polynomials only. In the present article a new minimization condition is presented, which can be applied to a wider class of functions. Experience with solving numerical test examples is presented in the concluding part of the paper.