糖域模型

来自集智百科 - 复杂系统|人工智能|复杂科学|复杂网络|自组织
跳到导航 跳到搜索


糖域模型 Sugarscape models是一个遵循简单规则的基于主体的模型,用于人工智能基于主体建模以及社会仿真等领域,在约书亚·爱泼斯坦 Joshua M. Epstein和罗伯特·阿克斯泰尔Robert Axtell一起出版的书籍中《不断发展的人工社会》Growing Artificial Societies中有详细介绍。[1]

起源

糖域模型最早可以追溯到马里兰大学 University of Maryland的经济学家托马斯·谢林 Thomas Schelling发表的论文 ,他提出了谢林的种族隔离模型 Schelling's model of segregation[2]。这篇论文是谢林和社会环境建模的同事写于1969年,但由于缺乏足够的算力与合适的编程机制来充分开发其模型的潜力,他们的选择受到了限制。


Epstein和Axtell在他们的书中,应用了约翰·何顿·康威 John Horton Conway提出来生命游戏的主体仿真模型,并将其应用于谢林的最初的想法中。为了证明他们在基于主体仿真领域的发现,他们创造了一个模型并将其制成CD-ROM附在书上一起出售。这个模型就是我们现在众所周知的概念”糖域模型“[1] 。此后,糖域模型就被广泛使用于基于主体的模型中,只要这些模型使用的规则与Epstein&Axtell定义的规则相似,都可以被称作是”糖域模型“。

原理

所有的糖域模型都包括主体(居民),环境(二维网格)以及控制主体之间,主体和环境之间相互作用的规则。


The original model presented by J. Epstein & R. Axtell (considered as the first large scale agent model) is based on a 51x51 cell grid, where every cell can contain different amounts of sugar (or spice). In every step agents look around, find the closest cell filled with sugar, move and metabolize. They can leave pollution, die, reproduce, inherit sources, transfer information, trade or borrow sugar, generate immunity or transmit diseases - depending on the specific scenario and variables defined at the set-up of the model.

Epstein和Axtell最初提出来的模型(被认为是第一个大规模社会仿真的主体模型),是基于51x51的细胞网格,其中每个细胞可以包含不同量的糖(或香料)。在每个步骤中,主体都会环顾四周,找到离自己最近的充满糖的细胞网格,然后移动并代谢。它们可能会留下污染,死亡,繁殖,继承资源,传递信息,交易或借糖,产生免疫力或传播疾病-取决于模型设置时定义的特定情况和变量。

Sugar in simulation could be seen as a metaphor for resources in an artificial world through which the examiner can study the effects of social dynamics such as evolution, marital status and inheritance on populations.

模拟中的”糖“可以看作是人造世界中资源的隐喻,设计者可以通过该隐喻来研究社会动力学效应,诸如进化,婚姻状况和种群继承问题。[3]


Epstein和Axtell在他们的书中提到的原始规则在精确模拟时可能会出现问题[4],因为并不总是能够复现出《发展中的人工社会》中所提供的结果。


模型实现

糖域模型已经有多种实现方式,其中一些可以在开源软件中获得。


Ascape

An original implementation was developed in Ascape, Java software suitable for agent-based social simulation. The Sugarscape model remains part of the built-in library of models distributed with Ascape. 一个最初的应用实现是基于Ascape开发的。Java软件适合做基于主体的社会仿真。糖域模型仍然是Ascape内嵌库模型中的一部分。[5]

NetLogo has been used to build Sugarscape models. Three Sugarscape scenarios are included in the NetLogo Models Library: "Immediate Growback", "Constant Growback" and "Wealth Distribution". Besides these three scenarios lies Iain Weaver's Sugarscape NetLogo model, which is part of the User Community Models Library. "It builds on Owen Densmore's NetLogo community model to encompass all rules discussed in Growing Artificial Societies with the exception of the combat rule (although trivial to include, it adds little value to the model)."[6] The model is equipped with rich documentation[7] including instructions for successful replication of the original Sugarscape rules.[4]

NetLogo已经可以实现Sugarscape模型。 NetLogo模型库中包含三个Sugarscape场景:“立即回升”,“恒定回升”和“财富分配”。除了这三种方案外,还有Iain Weaver的Sugarscape NetLogo模型,该模型是用户社区模型库的一部分。“它建立在Owen Densmore的NetLogo社区模型的基础上,涵盖了除战斗规则(尽管包含了琐碎的规则,但对模型却没有多大价值)之外的发展中的人工社会中讨论的所有规则。” 该模型配备了丰富的文档[7],其中包括成功复制原始Sugarscape规则的说明。[4]



SugarScape on steroids

Due to the emergent nature of agent-based models (ABMs), it is critical that the population sizes in the simulations match the population sizes of the dynamic systems being modelled.[8] 随着基于主体的模型(ABM)的兴起,在建模时,仿真中的总体大小与动态系统的总体大小相匹配是至关重要的。 However, the performance of contemporary agent simulation frameworks has been inadequate to handle such large population sizes and parallel computing frameworks designed to run on computing clusters has been limited by available bandwidth. As computing power increases with Moore's law, the size and complexity of simulation frameworks can be expected to increase. The team of R. M. D’Souza, M. Lysenko and K Rahmani from Michigan Technological University used a Sugarscape model to demonstrate the power of Graphics processing units (GPU) in ABM simulations with over 50 updates per second with agent populations exceeding 2 million. 但是,现代主体仿真程序框架的性能不足以处理如此大的人口规模,并且设计为在计算集群上运行的并行计算框架受到可用带宽的限制。根据摩尔定律,随着计算能力的增加,仿真框架的规模和复杂度也会随之增加。来自密歇根理工大学的R. M. D’Souza, M. Lysenko and K Rahmani使用糖域模型来证明ABM仿真中的图形处理单元(GPU)每秒更新50多次,主体人数超过200万。[9]

Mathematica

也可以在Mathematica上实现。[10]

MASON

乔治·梅森大学 George Mason University的MASON项目中,只要有免费学术许可证 Academic Free License,也可以实现糖域模型,其中MASON是使用Java开发的多主体仿真环境。 [11]

References

  1. 1.0 1.1 Epstein, Joshua M.; Axtell, Robert (October 11, 1996). Growing artificial societies: social science from the bottom up. Brookings Institution Press. pp. 224. ISBN 978-0-262-55025-3. https://archive.org/details/growingartificia00epst/page/224. 
  2. "Sugarscape - Growing Agent-based Artificial Societies". Sourceforge. Retrieved 7 November 2010.
  3. "Agents at Work". CIO Insight. 1 (27): 43. 1 June 2003. ISSN 1535-0096. Retrieved November 11, 2010.(Retrieved from ABI/Inform Document ID: 347271391)
  4. 4.0 4.1 "Replicating Sugarscape — University of Leicester". Archived from the original on 2012-06-19. Retrieved 18 January 2011.
  5. "The Ascape Model Developer's Manual". Sourceforge. Retrieved 9 November 2010.
  6. "NetLogo User Community Models: Sugarscape". Retrieved 9 November 2010.
  7. "The Sugarscape". University of Leicester. Archived from the original on 2017-10-02. Retrieved 19 January 2011.
  8. Gilbert, Nigel; Bankes, Steven (2002). "Platforms and Methods for Agent-Based Modelling" (PDF). Proceedings of the National Academy of Sciences. 99 (3): 7197–7198. doi:10.1073/pnas.072079499. PMC 128584. PMID 12011398.
  9. D'Souza, Roshan M.; Lysenko, Mikola; Rahmani, Keyvan (2007). "SugarScape on steroids: simulating over a million agents at interactive rates" (PDF). Proceedings of Agent2007 Conference. Chicago, Il.(See also: presentation slides)
  10. "Sugarscape: Agent-Based Modeling - Wolfram Demonstrations Project". Wolfram. Retrieved 18 January 2011.
  11. Bigbee, Anthony; Cioffi-Revilla, Claudio; Luke, Sean (2007). Terano, T.; Kita, H.; Deguchi, H.; et al. (eds.). "Replication of Sugarscape Using MASON" (PDF). Agent-Based Approaches in Economic and Social Complex Systems IV: Post-Proceedings of the AESCS International Workshop 2005. Tokyo: Springer.

External links