Exploring the evolution of internal control structure using digital enzymes
Exploring the evolution of internal control structure using digital enzymes
复制标题
使用数字酶探索内部控制结构的演变
DOI：
10.1145/2330784.2330956
复制
发表时间：
2012
影响因子：
14.3
通讯作者：
P. McKinley
中科院分区：
文献类型：
--
作者：
Chad M. Byers;B. Cheng;P. McKinley
控制的数字酶模型是基于细胞中发现的自下而上的信号转导的反应过程。早期的一项研究将这一模型的一个具体实例应用于觅食问题。在这里，我们扩展了该系统，并用它来探索生物学和进化计算中的一个基本问题，即环境复杂性是否是生物体S内部控制结构的驱动因素。为了解决这个问题，我们对原始系统进行了扩展，以允许每个控制器中独特的程序、指令和线程的开放式演变。有了扩展的模型，我们能够进化出成功的觅食策略，这些策略的性能几乎是早期工作中发现的策略的两倍。作为对日益增加的环境复杂性的响应，我们发现产生成功策略的程序、线程和指令的数量高度不同。这些结果表明，环境的复杂性不需要进化搜索方法来探索搜索空间中具有并行和分布式控制特征的区域。然而，在这些区域内发现的战略与由单一程序和主线管理的战略一样成功，突显了进化搜索技术的重要性，这使得关键内部控制组成部分能够无限期地演变。
The Digital Enzyme model of control is based on the bottom-up, reactive process of signal transduction found in cells. An earlier study applied a specific instance of the this model to the foraging problem. Here, we extend the system and use it to explore a fundamental question in both biology and evolutionary computation, namely, whether environmental complexity is a driving factor for an organism''s internal control structure. To address this question, we extended the original system to allow the open-ended evolution of the unique programs, instructions, and threads within each controller. With the extended model, we were able to evolve successful foraging strategies that nearly doubled the performance of strategies found in the earlier work. In response to increasing environmental complexity, we discovered a high degree of variation for the number of programs, threads, and instructions that produced successful strategies. These results imply that environmental complexity does not require evolutionary search methods to explore regions of the search space characterized by parallel and distributed control. However, strategies found within these regions were as successful as strategies governed by a single program and thread, highlighting the importance of evolutionary search techniques that enable the open-ended evolution of key internal control components.