Improved strategies for radial basis function methods for global optimization (Q868637): Difference between revisions

The global optimization method based on the radial basis function model has been proposed by \textit{H.-M. Gutmann} in 2001 [J. Glob. Optim. 19, No. 3, 201--227 (2001; Zbl 0972.90055)] for computationally expensive multimodal objective functions of modest dimensionality. In some cases the convergence of this method to the global minimum is slow. Two modifications proposed in this paper aim to enhance the efficienty of the original method. The first modification restricts the search region of the current iteration. In this way the authors want to ensure better balance between local and global search. The second modification defines a complete restart strategy when progress is slow during a preset number of iterations. Testing results are presented to illustrate the achieved improvements.