Research On Cultural Algorithm Based On Cellular Space Structure

Cultural algorithm is a new global optimization search algorithm, which reflectsthe evolution of the species accurately by simulating the evolution of human society, aswell as the evolution in the level of micro and macro. The base of traditional populationevolution, cultural algorithm can save all kinds of information in the form of knowledgein belief space. The information is created during the process of population evolution.Then all kinds of knowledge to react on population space, to guide the evolution of theindividual operations, so as to improve the efficiency of the population evolution,quickly achieve the optimal solution. With the emergence and development of culturalalgorithms, because of its good adaptability, robustness, usability and ease of operation,it has been get more and more attention and applied to various areas. To solve functionoptimization problems, cultural algorithm can also blend in different model structure,and can mix other ideas of optimization algorithm, and be able to make up for thedeficiency of the traditional algorithm, the convergence speed and global convergenceability shows a better effect than other similar algorithms.For making it more realistically simulate the human living environment, a morecomplicated geographical spatial system, highlighting the evolution mechanisms of thesociety and culture, to accelerate the speed of the population evolution, to furtherimprove the convergence accuracy and global convergence performance. The structureof cellular space is introduced into the culture algorithms in this paper, and theinformation of individual and evolution are stored in the cellular space respectively. Inpopulation space, establish geographic space and into other evolutionary algorithms tosimulate the evolution of the population, ensuring a diversity of individual. In beliefspace, this paper propose a new mechanism of cultural evolution, and establishing acultural diffusion model in cellular space structure, and using a variety of knowledgeintegration strategy, and expand the spread and diffusion theory to realize the culturalevolution in different areas. The main research content is as follows:(1) This paper introduces the basic principle of cultural algorithms. Each part ofcultural algorithm is described in detail. Three different test functions are selected to testcultural algorithm, cellular genetic algorithm and differential evolution algorithmrespectively. It proved that the cultural algorithm in global convergence ability and commonality has advantages, but had the slow convergence and low convergenceprecision when solved high-dimensional complex optimization problem.(2) In order to improve the performance of culture algorithm. This paper introducethe cellular space structure in the basic framework of culture algorithm, By usingdifferential evolution and the evolution cellular genetic algorithm based on populationdensity in population space, the evolution of individual is conducted, which haschanged simple rules of the past. It had both ensured individual populations of diversepopulations in space and accelerated the evolution of the individuals. The optimizingperformance of the algorithm is improved. Six typical test functions are used to testimproved algorithm, common cultural algorithm, cellular genetic algorithm, geneticalgorithms with evolution rules and differential evolution algorithm respectively, andthe optimizing performances of each algorithm are analyzed. The test results show thatthe improved algorithm can improve the accuracy of the algorithm of optimizationeffectively. It has reduced the convergence time and improved the stability of thealgorithm.(3) In order to solve those problems as the population space without the concept ofregion, the lack of cultural evolution in the belief space, and some defects like the lowaccuracy and some aspects are easy to fall into the local optimum when solving theoptimization problem in previous cultural algorithm. It can more realistically simulatethe human living environment and the complicated geographical spatial system byembedding cellular space grid structure in the computing framework of populationspace and belief space. For the population space, the evolutionary individuals aredistributed in the lower cellular space grid and the grid is divided into many areas, thusindividuals in each area evolve independently by using differential evolution algorithm.For the belief space, the evolution information is put into the upper grid correspondingto the population space, and the evolution of culture is realised by using diffusionmechanism of culture and knowledge fusion strategy. The results of thisexperiment show that the algorithm is not only effective in convergence accuracy andglobal search capability, but also has advantages when dealing with the complexhigh-dimensional optimization problems.