Gradient-Controlled, Typical-Distance Clustering for Global Optimization

I.G. TSOULOS (sheridancs.uoi.gr)
I.E. LAGARIS (lagariscs.uoi.gr)

Abstract: We present a stochastic global optimization method that employs a clustering technique which is based on a typical distance and a gradient test. The method aims to recover all the local minima inside a rectangular domain. A new stopping rule is used. Comparative results on a set of test functions are reported.

Keywords: Stochastic Global optimization, Multistart, Stopping rules

Category 1: Global Optimization

Category 2: Global Optimization (Stochastic Approaches )

Citation: Preprint, no 4-5/2004 Dept. of Computer Science, University of Ioannina, Greece

Download: [Postscript][PDF]

Entry Submitted: 05/17/2004
Entry Accepted: 05/17/2004
Entry Last Modified: 05/17/2004

Modify/Update this entry