Abstract:

A new proposal for multipoint rotate-vector optimization is presented based on the basic rotate-vector method. Some measures are adopted as increasing diversity of initial points, introducing rejection threshold and retained coefficient, then the capacity of global optimization and computational efficiency are improved by solving the examples. Genetic algorithm is compared with this method from points of view of result of example and principle of algorithm. The influence of every parameter is analyzed on the computational accuracy and time of multipoint rotate-vector method. The directory evaluation is given which is helpful to parameter setting in practical application.

Key words: rotate-vector, global optimization, genetic algorithm