A new filled function method for nonlinear integer programming problem (Q2489450)

Many approximate algorithms have been developed for solving the following problem belonging in general to the class of NP-hard problems: Minimize \(f(x)\) subject to \(x\in X_Z\), where \(X_Z\) is a given set of integer points in \(\mathbb R^n\). Some of these methods are the filled function methods, which are connected with the following two difficulties: 1) they do not guarantee the existence of local minimizers over \(X_Z\) except the vertices of \(X_Z\), 2) they have two parameters dependent on each other, which results in some difficulties in the process of appropriate adjustment of the parameters in such a way that they satisfy the required conditions. In the present paper, the authors propose a new filled function with only one parameter \(r\), which keeps the properties of the existing filled functions and ensures at the same time that a feasible point is a lower local minimizer of the original optimization problem if and only if it is a local minimizer of this filled function except the vertices of \(X_Z\) when the parameter \(r\) is sufficiently small. Then they can obtain for a sufficiently small \(r\) a lower local minimizer of the original optimization problem just by solving the problem with the filled function via local search schemes. The results obtained by solving several illustrative numerical examples are contained in the concluding section of the paper.