Experiments with New Stochastic Global Optimization Search Techniques

Abstract

In this paper several probabilistic search techniques are developed for global optimization under three heuristic classifications: simulated annealing, clustering methods and adaptive partitioning algorithms. The algorithms proposed here combine different methods found in the literature and they are compared with well-established approaches in the corresponding areas. Computational results are obtained on 77 small to moderate size (up to 10 variables) nonlinear test functions with simple bounds and Is large size test functions (up to 400 variables) collected from literature.