An interval algorithm for nondifferentiable global optimization (Q1335231)

Let \(G\) be the set of global solutions of the constraint optimization problem \(\min f(x)\), \(x\in S\), and let \(G_ c\) be the set of global solutions of the unconstrained problem \(\min F(x)\) with \(F= f+ cp\), where \(p\) is a penalty term for the feasible set \(S\) and \(c\) is a penalty factor. In the theory of penalty functions, it is well-known that the accumulation points of any sequence \((x_ c)\) with \(x_ c\in G_ c\) lie in \(G\) as \(c\to\infty\). The authors show now that under certain assumptions a number \(\bar c\) can be constructed such that \(G_ c= G\) for all \(c\geq \bar c\). This property combined with interval arithmetic techniques is used to propose an effective deterministic algorithm for solving the problem originally given. A very detailed PASCAL code and numerical results are added.