• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>IEEE Symposium Series on Computational Intelligence >A preliminary study on designing a benchmark problem for analysis of sparsely-synchronized heterogeneous coevolution
【24h】

A preliminary study on designing a benchmark problem for analysis of sparsely-synchronized heterogeneous coevolution

机译：设计基准问题以分析稀疏同步异构进化的初步研究

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Cooperative coevolution evolutionary algorithms (CCEAs) that divide a target problem into subproblems and optimize them using multiple component solvers have attracted attention in recent years because they solve large-scale optimization problems. On the other hand, sparsely-synchronized heterogeneous coevolution (SSHC) which solves subproblems with different characteristics has also been proposed and expected to optimize various problem combinations that have not been solved simultaneously so far. SSHC simultaneously optimizes subproblems that depend on each other, and sometimes requires different types of component optimizers due to the difference in their properties, solution structures, etc. Unlike general CCEAs, SSHC algorithms cannot dynamically change problem decomposition way during optimization. Then, there is concern that inter-subproblem dependency deeply affects SSHC performance. In general, benchmark problems are indispensable to investigate algorithm dynamics and behavior. However, no benchmark problems for cooperative coevolution in continuous domain that allow changing dependency strength have been proposed so far. Therefore, this paper attempts to design a benchmark problem named Continuous NK Landscape for Cooperative Coevolution (CNKLcc), which allows arbitrarily changing inter-and intra-subproblem dependencies. A Continuous NK landscape problem is employed to design CNKLcc so that both dependency type and strength between subproblems can be controlled. This paper also analyzes SSHC performance using CNKLcc to verify the effectiveness of their parameter adjustment. Experimental results showed that the performance of SSHC algorithms would be improved by devising appropriate collaboration model as with general CCEAs.
机译：近年来，合作协作进化算法（CCEA）将目标问题分解为子问题，并使用多组件求解器对其进行优化，这是因为它们可以解决大规模优化问题，因此备受关注。另一方面，也提出了解决具有不同特征的子问题的稀疏同步异构协进化（SSHC），并有望优化到目前为止尚未同时解决的各种问题组合。 SSHC同时优化相互依赖的子问题，由于其属性，解决方案结构等方面的差异，有时需要使用不同类型的组件优化器。与常规CCEA不同，SSHC算法无法在优化过程中动态更改问题分解方式。然后，人们担心子问题间的依赖性会严重影响SSHC的性能。通常，基准问题是研究算法动态和行为必不可少的。但是，到目前为止，尚未提出允许连续变化的依赖强度的连续域中协作协同进化的基准问题。因此，本文尝试设计一个基准问题，称为协作协同进化的连续NK景观（CNKL cc ），该问题允许任意更改子问题之间和子问题之内的依赖性。为了设计CNKL cc ，使用了连续NK景观问题，以便可以控制依赖关系类型和子问题之间的强度。本文还使用CNKL cc 分析SSHC性能，以验证其参数调整的有效性。实验结果表明，通过与通用CCEA一起设计适当的协作模型，可以提高SSHC算法的性能。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号