Numerical comparisons of migration models for Multi-objective Biogeography-Based Optimization
Numerical comparisons of migration models for Multi-objective Biogeography-Based Optimization
复制标题
基于多目标生物地理学的优化迁移模型的数值比较
DOI：
10.1016/j.ins.2015.07.059
复制
发表时间：
2016
影响因子：
8.1
通讯作者：
Wu Qidi
中科院分区：
文献类型：
--
作者：
Guo Weian;Wang Lei;Wu Qidi
近年来，基于生物地理的优化算法（BBO）作为一种新颖的进化算法（EA）因其优异的性能而备受关注。在我们之前的工作中，研究了BBO的单目标问题迁移模型，揭示了它们对算法性能的影响。然而，目前对于多目标问题（MOPs）的迁移模型研究较少，这类问题在实践中较为常见，但难度较大。为了使BBO在处理MOPs方面具有竞争力，本文对BBO的迁移模型进行了探索和开发。我们的工作贡献之一是提出了多目标BBO （MOBBO）。将MOBBO算法与其他流行的moea算法进行比较，表明该算法能够处理moops。此外，我们提出并比较了六种主要的迁移模型。通过比较，我们发现梯形迁移模型对MOPs的迁移性能较好，而对sop的迁移性能不如其他迁移模型。此外，二次迁移模型对MOPs的求解性能较差，而对sop的求解性能较好。本文的另一个贡献在于，我们从经验的角度重新评估了MOPs的迁移模型，这有助于设计MOBBO的迁移模型。
In current years, Biogeography-Based Optimization (BBO), a novel Evolutionary Algorithm (EA), has drawn a lot of attention due to its dramatic performance. In our previous work, BBO’s migration models for single-objective problem (SOP) have been investigated to reveal their effects to algorithm’s performance. However to date, there is a few investigation about migration models for Multi-Objective Problems (MOPs) which are common in practice though more difficult. To make BBO competitive in dealing with MOPs, migration models of BBO are explored and exploited in this paper. One contribution of our work is that we propose Multi-objective BBO (MOBBO). By comparing MOBBO with other popular MOEAs, this algorithm is competent to handle MOPs. Besides, we present and compare six principal migration models. In comparisons, we find that Trapezoidal Migration Model performs well for MOPs, while its performance is inferior to other migration models for SOPs. Besides, Quadratic Migration Model’s performance for MOPs is worse, while it has a good performance to solve SOPs. These demonstrate that the conclusion to evaluate migration models for SOPs does not hold for MOPs, so another contribution in this paper is that we reevaluate migration models for MOPs in an empirical way, which is helpful to design migration models for MOBBO.