This thesis has described an optimization method that was designed to be robust and efficient. The ideas that built the method came from an inquiry into the nature and use of optimization programs based on genetic algorithms. The method uses information about the discovered local maxima to guide the search. The step sizes used are adaptive to the local terrain of the search space. Finally, the direction of most rapid ascent is computed from recent successful moves.
The method was tested on several problems and compared to other successful optimization algorithms. It has actively improved the solutions for real world problems selected from science, medicine and engineering. Its advantage is demonstrated on problems with high number of dimensions and many local maxima.
Theoretical work on the assumptions and performance of the method is required and the application to combinatorial problems need more study. Most importantly, more problems from different fields need to be found and studied in the context of the ideas presented in this thesis.