New methods for developing parallel algorithms of direct search in numerical optimization

Abstract<br /> The Purpose of this research is the development Direct Search Methods for minimizing the objective function; <br /> The Developed Methods reduces the approach time to the minimum through two techniques:<br /> 1) by reducing the number of function evaluation.<br /> 2) by developing algorithms suitable for running on MIMD(Multiple Instruction Stream Multiple Data Stream).<br /> We developed two methods, the first method depends on the "pattern search" and the other depends on the "direct search" technique. The comparison of results shows the new algorithms are faster than corresponding algorithms.