| In this paper, we study a new kind of evolutionary algorithm for global optimization over continuous spaces, a revision of Differential Evolution based on piecewise two-dimensional search.For global optimization problem, we are always eager to know its solution as soon as possible and accurately. Because the optimization problem may be discontinuous and non-differentiable even has no objective function, the traditional optimization methods are unable to tackle with it. Evolutionary Algorithm is a kind of self-organizational and adaptive stochastic search algorithm which is advanced based on natural selection and Darwin's main principle: survival of the fittest. It employs heuristics which enhances the search, so it is popular in many fields. Rainer Storn and Kenneth Price put forward a simple and efficient adaptive scheme for global optimization over continuous spaces-Differential Evolution. DE, simulating nature and iterating according to probability, belongs to a kind of stochastic search algorithm. It's superior to some of the traditional EAs, however, sometimes it converges slowly and might be premature. Sometimes it is costly to compute objective function of the optimization problem which makes DE infea-sible. In order to solve this problem, we propose new evolutionary scheme for partitioning the searching field. With this method, we can search a nearby planar space in searching fields around each vector in present generation. To some extent, this method makes search with determination. This property could accelerate convergence. We also test a lot of optimization problems to support our algorithm. The fact shows that this method is feasible and efficient because it can reduce times of the computing objective function while the solution's accuracy is good. That is to say we found the speed is faster than before. A lot of valuable experience about how to apply evolutionry algorithm into some practical problems have been gotten from this research.
|
PDF Full Text Request