The Study Of Filled Function On Global Optimization Problems

Optimization as the practicability of a discipline, mainly discuss how to use theexisting conditions, choose the best decision-making to achieve the best effect, so as tofind the optimal solution of original problem. Global optimization problem as animportant branch of optimization, it is common in economic, scientific, the production,transportation planning, military operations, and other fields, the global optimizationproblems by the majority of scientific and technical workers, has a rich theoreticalknowledge and algorithm. The method to solve the problem is divided into twocategories, One is the deterministic method; Second is the stochastic method. the filledfunction method as a kind of effective deterministic method, on the one hand, it canjump out of local optimal solution, to find the global optimal solution; On the otherhand, filled function is the objective function of the composite function, therefore, thefunction will not only affected by the objective function, but also affects the precision ofthe algorithm can find the optimal solution and the solution precision. Therefore,Structure has good properties, simple structure of the filling function, is to further to digup the potential of filling function, is also the theoretical and practical workers continueto study a problem.In this paper, the research has the following several aspects:1. Through the analysis of the existing filling function, found the followingdeficiencies is poor, for example filling function properties is poor, parameters are moredifficult to adjust; Second, assume the objective function has some properties, such asmeet mandatory, Lipschitz condition. To this, we consider the objective functionwithout Lipschitz condition, and construct a filled function with a parameter, discuss therelated properties of the function, design The corresponding algorithm, through thenumerical experiment, the results show that the method is feasible.2. As the filled function with parameters in numerical experiments has difficulty, ifthe parameter adjustment is undeserved, can affect the result of the experiment, at thesame time, reduce the efficiency of algorithm. To this, we construct a filled functionwithout parameter, and analyzes the characters of the function, design The correspondingalgorithm, through the numerical experiment, the experimental results show that thismethod is feasible.3. Further analysis the existing no parameter function, found the function is not including the objective function. To this, we construct a new filled function parameters,and analyzes the characters of the function, design The corresponding algorithm, through thenumerical experiments, the results show that the method is effective.