From Genetic Programming to Computational Evolution

从遗传编程到计算进化

• 批准号: 283304-2012
• 负责人: Banzhaf, Wolfgang
• 金额: $ 2.48万
• 依托单位: Memorial University of Newfoundland
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2015
• 资助国家: 加拿大
• 起止时间: 2015-01-01 至 2016-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=571333
• 关键词: Genetic; Programming; Computational; Evolution

The goal of the research proposed here is to move from earlier very abstract and simplified EC models, closer to realistic models of natural evolution implemented in computational systems, in other words, in the direction of "computational evolution". To this end, a number of basic assumptions of established evolutionary algorithms will be subjected to critical revision. In addition, new concepts of molecular and evolutionary biology not known at the time of the design of these algorithms will be incorporated. While the set of those concepts is virtually endless, my focus for the next five years is on the following topics: (i) epigenetic interactions that allow a genome to reorganize in response to cues from its environment. This is an extremely important concept, notably with respect to the reinforced connection between evolution and development in biology. The inclusion of epigenetic mechanisms in algorithms will happen through the addition of a further layer of control signals for artificial regulatory networks. This constitutes a new way of allowing interactions between an algorithmic genome and its environment. (ii) The evolution of cooperation and its contribution to evolutionary transitions. While earlier models studied two-level systems, in this research we shall study multi-level selection and the benefits to be gained from adapting the size of groups and hierarchies to the problem that is to be solved by evolution. Substantial efficiency gains are expected from these approaches, since they address one of the key problems evolutionary algorithms, and Genetic Programming in particular: scalability. (iii) The emergence of novelty in both biological and computational environments. Novelty and the associated phenomenon of innovation is a central tenet in Darwinian thought. Without the invention of new devices neither nature nor computing reap the full benefits of evolution. Here, we shall examine the mechanisms by which novelty is produced in natural settings, for example by gene duplication in genomes, and seek to transfer recipes into computational systems.

这里提出的研究目标是从早期非常抽象和简化的EC模型转向计算系统中实现的自然进化的现实模型，换句话说，朝着“计算进化”的方向发展。为此，已建立的进化算法的一些基本假设将受到严格修订。此外，还将纳入在设计这些算法时未知的分子和进化生物学的新概念。虽然这些概念实际上是无穷无尽的，但我未来五年的重点是以下主题： (i) 表观遗传相互作用，使基因组能够根据环境线索进行重组。这是一个极其重要的概念，特别是对于生物学中进化与发展之间的强化联系而言。通过为人工调控网络添加另一层控制信号，可以将表观遗传机制纳入算法中。这构成了一种允许算法基因组与其环境之间交互的新方式。 (ii) 合作的演变及其对演变转变的贡献。虽然早期的模型研究的是两级系统，但在本研究中，我们将研究多级选择以及通过调整群体和层次结构的大小来适应进化要解决的问题所获得的好处。这些方法预计会带来显着的效率提升，因为它们解决了进化算法，特别是遗传编程的关键问题之一：可扩展性。 (iii) 生物和计算环境中出现新颖性。新颖性和相关的创新现象是达尔文思想的核心原则。如果没有新设备的发明，自然和计算都无法获得进化的全部好处。在这里，我们将研究自然环境中产生新颖性的机制，例如通过基因组中的基因复制，并寻求将配方转移到计算系统中。