Svetlana Stepanenko

Global Optimization Methods Based on Tabu Search: Gts, Gots, Tspa. Application for Conformation Searches. Svetlana Stepanenko

Many activities can be formulated as optimizationproblems. Because of the computational complexity ofthese problems, exact optimization techniques ofoperations research like linear programming ordynamic programming are for the most part unfeasiblefor large-scale problems. Therefore, researchers,engineers, and scientists have used metaheuristicsearch techniques to find near optimal, good-enoughoptimal or even the global solutions. This work encompasses three parts. The first partprovides a concise review of the most prominentmetaheuristic concepts currently available and givesessential preliminaries together with definition ofthe combinatorial optimization problems. In thesecond part the new nonlinear global optimizationroutines based on the TS strategy are described: theGradient Tabu Search (GTS), the Gradient Only TabuSearch (GOTS), and the Tabu Search with Powell?sAlgorithm (TSPA). The new algorithms are comparedwith other approaches using various well-known testfunctions with varying dimensionality. In the lastpart of the work the GOTS is applied for globalminimum energy conformation.