Estimation of Distribution Algorithms with Graph Kernels

Estimation of Distribution Algorithms with Graph Kernels
复制标题

使用图核的分布算法的估计

DOI:
10.11394/tjpnsec.7.56
复制
发表时间:
2016
期刊:
Transaction of the Japanese Society for Evolutionary Computation
影响因子:
--
通讯作者:
前澤健太,半田久志
前澤健太,半田久志
中科院分区:
--
文献类型:
--
作者:
芳野 裕規;近藤 秀明;坪 泰宏;前澤健太,半田久志

文献摘要

相似文献

在本文中,我们提出了一种新的进化算法来解决个体用图表示的问题。为了解决个体基因型-表型映射的困难,我们将图核的概念纳入分布估计算法。也就是说,在该方法中,个体的接近度不是在基因型空间上定义的,而是在特征空间上定义的。我们在Edge-Max, Edge-Min和图同构问题的几个实验中证明了该方法的有效性。
In this paper, we propose a novel evolutionary algorithm for solving problems such that individuals are represented by graphs. In order to address the difficulty of genotype-phenotype mappings of individuals, we incorporate a notion of Graph Kernels into Estimation of Distribution Algorithms. That is, the proximity of individuals in the proposed method is defined not on genotype space but on feature space. We show the effectiveness of the proposed method on several experiments on Edge-Max, Edge-Min, and graph isomorphic problems.