Parallel Optimization Methods Based on Direct Search

摘要: This paper is focused in the parallelization of Direct Search Optimization methods, which are part family derivative-free methods. These methods known to be quite slow, but easily parallelizable, and have advantage achieving global convergence some problems where standard Newton-like (based on derivatives) fail. been tested with Inverse Additive Singular Value Problem, a difficult highly nonlinear problem. The results obtained compared those derivative methods; efficiency parallel versions has studied.