The Research Of Diversity On Multi-Objective Evolution Algorithm

A Multi-Objective Problem has to calculate more than one problem at the same time of witch conflict with eachother. To solve those multi-objective problems, researchers have developted some Multi-Objective Evolutionary Algorithms. Dury calculating and evoluting the multi-objective problem, multi-objective evolutionary algorithm will get sorts of non-dominate Pareto solutions. Beacause Multi-Objective Problem in our wold always non-liner and even linkage between each variable, it is difficult for traditional methods to find out non-dominate Pareto solution sets. Multi-Objective Evolutionary Algorithms generate new solutions depends on faher population with some selection strategy. In order to make press for the population to evolate to Pareto Front, Multi-Objective Evolutionary Algorithm will keep the non-dominated individuals with good diversity. The steps will stop until computer resourse or condition was achieved.The population’s distribution is very important when assessing the performance of a multi-objective evolutionary algorithm. Maintain the population distribution is not only related to the decision-makers be able to get complete information, but also a important indicator when evaluate a multi-objective evolutionary algorithm. Population distribution studies including populations in the object space distribution as well as in the decision space distribution. Because it is difficult to find the right starting point, multi-objective evolutionary algorithm distribution study only focused on the target space, there is little system analysis and research for objective space and decision space.This paper treat niche as the starting point to research the object space and decision space distribution of multi-objective evolutionary algorithm. Niche in the object space simulation animals in their natural environment to keep distribution of the population, and take advantage of the minimum spanning tree to adjust niche radius. In the decision space, we proposed a stratagem of individual choice based on niche strategy, this strategy can take advantage of the individual diversty information in decision space to adjust the search direction. This statage can improve the search efficiency of multi-objective optimization problem with variable linkage.