Genetic Algorithm Research And Application

The genetic algorithm is a kind of searching method using probability which simulates the natural evolution. Its predominance lies in effective resolving complicated and non-linear problems, which are difficult for traditional searching methods. Its few limitation on the presumption of the solution space, wide adaptability and parallelism make it successfully applied in many area, including function optimizations, machine learning, pattern recognition and self-adaptive control systems, etc.Firstly, the paper briefly summarizes the rationales of GA. Then, two new modified strategies have been proposed. They are used to solve tow problems which are premature convergence and blind genetic operators. In the first method, an index that can reflect the orientation and pace of evolution is defined. The index is used to guide the evolution and to adaptively adjust genetic operator. Based on a multi-population scheme consisting of multi-explorer and an exploiter, a parallel edition of the method is designed and tested on the Cluster of computer.The second method presents a kind of hybrid GA based on particle swarm algorithm. It defines a new multi-population scheme and reconstructs the mutation operator. Further more, the recombination operator is also improved. Three typical multimodal functions are optimized and the optimization results are analysed. The paper still solves equations with the hybrid algorithm, and the experimental results show that it can effectively find all roots of equations.Finally, the paper points out the weak points, limitation of the two strategies and makes a plan for further study.