Speed Up Reliability Model Optimization With Hypervolume Contribution Calculating Algorithm
Speed Up Reliability Model Optimization With Hypervolume Contribution Calculating Algorithm
复制标题
DOI：
10.1080/10798587.2011.10643175
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
Xiuling Zhou;Ping Guo;C. L. P. Chen
中科院分区：
文献类型：
--
作者：
Xiuling Zhou;Ping Guo;C. L. P. Chen
摘要软件可靠性建模需要同时考虑多个不相容且经常相互冲突的目标，而基于超体积的多目标进化算法（MOEA）在实际应用中对多目标问题具有更好的效果。提出了一种基于超体积的MOEA可靠性模型优化框架。针对基于超体积的MOEA算法的关键问题，提出了一种新的基于切片目标的集超体积贡献算法（SHSO），该算法在小维情况下直接计算子集对整个非支配集的唯一超体积贡献.对于SHSO的特殊情况，利用几何学方法对CHSO（切片目标对超体积的贡献）进行了改进。实验结果表明了所提出算法的可行性和有效性。
Abstract Software dependability modelling involves simultaneous consideration of several incompatible and often conflicting objectives, while hypervolume-based multi-objective evolutionary algorithm (MOEA) has been shown to produce better results for multi-objective problem in practice. A frame of reliability model optimization with hypervolume based MOEA is presented. Focusing on the key issue of hypervolume based MOEA, a new algorithm, set hypervolume contribution by slicing objective (SHSO), is proposed for calculating the exclusive hypervolume contribution of a subset to the whole nondominated set directly for small dimension. For the special case of SHSO, CHSO (the contribution of a point to hypervolume by slicing objective) is improved with heuristics. The feasibility and efficiency of developed algorithms are shown by experiments.