当前位置: X-MOL 学术Memetic Comp. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Compressed representation for higher-level meme space evolution: a case study on big knapsack problems
Memetic Computing ( IF 4.7 ) Pub Date : 2017-10-31 , DOI: 10.1007/s12293-017-0244-3
Liang Feng , Abhishek Gupta , Yew-Soon Ong

In the last decades, a plethora of dedicated heuristic and meta-heuristic algorithms have been crafted to solve complex optimization problems. However, it is noted that the majority of these algorithms are restricted to instances of small to medium size only. In today’s world, with the rapid growth in communication and computation technologies, massive volumes of data are generated and stored daily, making it vital to explore learning and optimization techniques that can handle ‘big’ problems. In this paper, we take an important step in the aforementioned direction by proposing a novel, theoretically motivated compressed representation with high-level meme evolution for big optimization. In contrast to existing heuristics and meta-heuristics, which work directly on the solution space, the proposed meme evolution operates on a high-level meme space. In particular, taking knapsack problem as the case study, a meme, in the present case, represents a knowledge-block as an instruction for solving the knapsack problem. Since the size of the meme, as defined in this paper, is not strongly sensitive to the number of items in the underlying knapsack problem, the search in meme space provides a compressed form of optimization. In order to verify the effectiveness of the proposed approach we carry out a variety of numerical experiments with problem sizes ranging from the small (100 items) to the very large (10,000 items). The results provide strong encouragement for further exploration, in order to establish meme evolution as the gold standard in big optimization.

中文翻译:

模因空间高级演化的压缩表示:以大背包问题为例

在过去的几十年中,为解决复杂的优化问题,设计了许多专用的启发式算法和元启发式算法。但是,应注意的是,这些算法中的大多数仅限于小到中等大小的实例。在当今世界中,随着通信和计算技术的迅速发展,每天都会生成和存储大量数据,因此探索能够处理“大”问题的学习和优化技术至关重要。在本文中,我们朝着上述方向迈出了重要的一步,提出了一种新颖的,具有理论动机的压缩表示形式,该压缩表示形式具有高级模因演化功能,可进行大的优化。与直接在解决方案空间上工作的现有启发式方法和元启发式方法相反,拟议的模因进化是在高级模因空间上进行的。特别是,以背包问题为例,在本案例中,模因代表知识块,作为解决背包问题的指令。由于本文定义的模因大小对基本背包问题中的项数不很敏感,因此在模因空间中搜索提供了一种优化的压缩形式。为了验证所提出方法的有效性,我们进行了各种数值实验,问题规模从小(100件)到超大(10,000件)不等。研究结果为进一步推动模因进化成为大优化的黄金标准提供了强有力的鼓励。表示一个知识块,作为解决背包问题的指令。由于本文定义的模因大小对基本背包问题中的项数不很敏感,因此在模因空间中搜索提供了一种优化的压缩形式。为了验证所提出方法的有效性,我们进行了各种数值实验,问题规模从小(100件)到大(10,000件)不等。研究结果为进一步推动模因进化成为大优化的黄金标准提供了强有力的鼓励。表示一个知识块,作为解决背包问题的指令。由于本文定义的模因大小对基本背包问题中的项数不很敏感,因此在模因空间中搜索提供了一种优化的压缩形式。为了验证所提出方法的有效性,我们进行了各种数值实验,问题规模从小(100件)到超大(10,000件)不等。研究结果为进一步推动模因进化成为大优化的黄金标准提供了强有力的鼓励。在模因空间中的搜索提供了优化的压缩形式。为了验证所提出方法的有效性,我们进行了各种数值实验,问题规模从小(100件)到超大(10,000件)不等。研究结果为进一步推动模因进化成为大优化的黄金标准提供了强有力的鼓励。在模因空间中的搜索提供了优化的压缩形式。为了验证所提出方法的有效性,我们进行了各种数值实验,问题规模从小(100件)到大(10,000件)不等。研究结果为进一步推动模因进化成为大优化的黄金标准提供了强有力的鼓励。
更新日期:2017-10-31
down
wechat
bug