动态多目标优化进化算法关键问题研究及应用

项目名称： 动态多目标优化进化算法关键问题研究及应用

项目编号： No.60803095

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 金属学与金属工艺

项目作者： 郑波尽

作者单位： 中南民族大学

项目金额： 18万元

中文摘要： 本项目对动态多目标优化问题和动态多目标优化演化算法进行了深入研究. ①#21462;得了重大进展: a.研究了带随机变量的多目标优化问题以及算法收敛行为,并将其应用到为复杂网络的结构起源建模,从而理论上澄清了各种复杂网络之间错综复杂的关系; b.只需改变该模型的参数,即可生成具有各种幂指数的无标度网络,社区网络,小世界网络,分形网络,类星型分布网络,随机网络等多种多样的网络,为现有的复杂网络提供了统一的起源解释,有望修正该领域中多个主流观点,如:无标度网络的起源,无标度网络在选择性攻击下的鲁棒性,分形网络的起源等问题,并提出新观点,如:社区网络的起源等问题; c.该模型有望用于解释如地震预测、禽流感传播、财富分配等多种关乎国计民生的重大问题. ②#21462;得了一定成绩:a.提出并完善了参考点方法,即几何Pareto选择文档算法,证明了该算法在二维目标条件下是最优文档算法,提出了相应的多目标优化演化算法,实验表明效果良好;b.探讨了动态优化演化算法;c.研究了基于超变异的动态多目标优化演化算法,并针对汽车外形智能设计这一应用进行改进,提出了动态交互式式多目标优化演化算法,能自适应地改进汽车的外形设计.

中文关键词： 动态多目标优化;演化算法;复杂网络;复杂系统

英文摘要： This project has deeply explored the dynamic multi-objective optimization problems and the dynamic multi-objective optimization evolutionary computation. ①he great advances: a. Explored the multi-objective optimiztion problems and the convergent behaviors of the algorithms and applied the results to model the origins of the structures of complex networks, so that this project clarified the complicated relationships between various complex networks by the theoretical way; b. With the changes of the parameters, the proposed algorithm can generate scale-free networks with arbitrary exponent, community-structured networks, small-world networkd, fractal networks, star-alike networks, random networks and so on, so that this project provided a universal explanation to the origins of complex networks. It is hopeful of modifying multiple mainstream viewpoints, such as the origin of scale-free networks, the robustness of scale-free networks under the selective attacks, the orgins of fractal networks and proposed some new viewpoints such as the origin of the community-structured networks; c. This model could be used to explain some important problems such as the prediction of the earthquakes, the spread of H1N1, the distribution of the fortune etc. ②ome results: a. Proposed and improved the reference point approach, i.e., the geometric Pareto Selection algorithm, and proved this algorithm is the best algorithm when the number of objectives is 2. Based on this algorithm, proposed the corresponding multi-objective optimization evolutionary algorithms. The experimental results are satisfactory. b. Explored the dynamic optimization evolutionary algorithms. c. Explored the dynamic multi-objective optimization evolutionary algorithms based on hyper-mutation and applied the results to the intelligent design of the shape of the cars. Proposed a dynamic multi-objetive optimization evolutionary algorithm, which is capable of evolving the shapes of the cars with self-adaption.

英文关键词： Dynamic Multi-Objective Optimization; Evolutionary Compuation;Complex Network; Complex Systems