An Improved Multi-niche Genetic Algorithm And Its Application

Multi-modal optimization problems widely spread in function optimization, design and decision-making, industrial applications and many other fields. As the genetic algorithm search process is carried out in a group of solution, the populations of individuals may be found in every corner of the solution space, so long as to maintain the population diversity, it is possible to search a number of global peak or local peak. Therefore, the genetic algorithm for solving multimodal optimization problems is a powerful tool.Niche genetic algorithms are the effective means to solve multimodal optimization problems. This paper analyzes the classic niche genetic algorithm, and cluster analysis will be introduced in multi-niche crowding genetic algorithm. An improved multi-niche crowding genetic algorithm is put forward, and this algorithm is applied to the optimization of shaped pulse in magnetic resonance imaging.In this paper, the following tasks are covered:1. The basic concept of genetic algorithms, basic principles and research status is briefly introduced.2. The basic idea, realization on niche genetic algorithm and typical niche genetic algorithm are introduced in detail. The advantages and disadvantages of various niche genetic algorithms are analyzed and contrasted.3. An improved multi-niche crowding genetic algorithm based on k-means clustering is proposed. Standard multi-niche crowding genetic algorithm has faster search rate and the relatively poor search capabilities. Though multi-niche crowding genetic algorithm with fitness-sharing can improve the search capabilities, the search speed is significantly dropped. To introduce cluster analysis can improve the choice and replacement mechanism. Numerical experiments show that the improved algorithm retains the speed advantage, and the search capability is also improved to some extent.4. The improved multi-niche crowding genetic algorithm is applied to the optimal design of shaped pulse in magnetic resonance imaging, and the special-shaped excitation pulse and reverse pulse are obtained. Compared with the simulated annealing and other methods, the improved multi-niche crowding genetic algorithm has the following advantages:(1) It can provide a variety of optimal design solutions to users with more choice; (2) It has a higher operating efficiency; (3) It has better global convergence.Figure 14 Table 6 Reference 41...