项目名称: 基于目标域分层的不确定高维多目标优化及其应用研究
项目编号: No.61300159
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 蔡昕烨
作者单位: 南京航空航天大学
项目金额: 23万元
中文摘要: 不确定高维多目标问题是智能优化领域面临的巨大挑战。现有的不确定多目标处理技术通过更改Pareto支配规则, 如Fuzzy dominance,对目标间的不确定性给予一定的容忍度,从而使优化算法具备较好的鲁棒性。然而,此类算法存在高维多目标、目标同等重要假设以及目标空间域结构关系的处理等诸多问题。本项目旨在利用先验的目标结构关系信息设计一种基于目标域分层的fuzzy dominance,并对此进行一系列的拓展研究。主要内容和目标为:1)通过先验的结构信息,将分层思想引入目标域空间的构造,设计基于目标域分层的fuzzy dominance,克服高维多目标等问题;2)将上述思想应用于优化算法的设计,设计一种新的不确定性高维多目标优化算法;3)应用上述算法于基于噪声数据的基因调控网络模型参数优化问题。
中文关键词: 高维多目标优化;帕里托前沿;不确定性;分解;排序
英文摘要: Uncertain many-objective problem is a major challenge in the field of intelligence optimization. Currently, most uncertainty handling techniques in multi-objective optimization, such as fuzzy dominance, are based on the modification of the Pareto dominance to give tolerance to the uncertainties when comparing solutions. In this way, the designed algorithm have the characteristics of robustness, to some level. Nonetheless, this type of algorithms have problems such as many objective, hypothesis of equal important objectives and the structural relation among objectives. This proposal aims to make use of structural information among objectives and design a relaxed form of Pareto dominance. The main content and objectives of this proposal are as follows: 1) use the concept of hierarchy in the construction of objective space through prior structural information and design the fuzzy dominance based on hierarchy of objectives, to overcome the many-objective, etc.; 2) extend the idea to the design of uncertain many-objective optimization algorithm; 3)apply the new algorithm to the parameters optimization of gene regulatory network models.
英文关键词: Many objective optimization;Pareto front;uncertainty;decomposition;sorting