Study On Swarm Intelligence Optimization Algorithm

Optimization algorithm based on the swarm intelligence is a simulated evolutionary method that simulating the behaviors of social insects searching for food and building of nest. At present, there are genetic algorithm, ant colony algorithm, particle swarm algorithm and artificial fish-swarm algorithm mainly. They are a kind of stochastic search iterative algorithm and no demand of optimization object. Because each algorithm has different mechanisms of search, different optimization characteristic and different application rage, they always depend on large numbers of experience and results of experimentations selecting suitable algorithm for factual optimization. In addition, the product of research in swarm intelligence optimization is separate and lack of systematization at the present time. There appear lots of similar algorithms. This phenomenon is of disadvantage to explore new synthesized mechanism optimization algorithm and exploit application rage. And it is negative for developing this field.Therefore, this paper will limit optimization object and only study function optimization, which is different from previous methods that only select several functions to study performance of optimization algorithm. A hill-valley algorithm is proposed which classify functions as hard monotony functions, hard unimodal functions, macroscopical monotony functions, macroscopical unimodal functions, and multimodal functions. Besides monotony functions and unimodal functions are classified as flat fitness landscape and rugged fitness landscape by the characteristic of function surface.By analyzing and discussing performance of operators and algorithms, this paper studies and compares all factors which affect the exploitation and exploration relationship (EER). It concludes the application rage and performance and efficiency of each optimization algorithm .Based on these a relatively perfect theory frame is proposed. The experiment proves conclusion of research.