• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International conference of the System Dynamics Society >Yet another Representation for System Dynamics Models, and Its Advantages
【24h】

Yet another Representation for System Dynamics Models, and Its Advantages

机译：系统动力学模型的另一种表示形式及其优势

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

摘要

The present paper relates to an artificial neural network (ANN) representation for asystem dynamics model (SDM) and its advantages in model construction and policy design.The similarities between SDMs and ANNs have been noted, i.e., both of which storeknowledge mainly in the structure (or linkages) of a model, rather than in the units or othercomponents. By a specially designed mapping scheme, it is shown that a given flow diagram(FD) (i.e., traditional representation for a SDM) can be mapped to a corresponding model inthe representation of partial recurrent networks (PRNs) that will correctly behave like the oneit mimics. Because a (partial recurrent) neural network can be trained with exemplar data,numeric propagation constraints can be identified by extracting rules from a set ofmultivariate time series of data by induction. This adds an advantage to the study of SD sinceit is made possible to create a model by learning instead of manual construction, which solelyrelied on experts’ observation and deduction quality. Similarly, it is also beneficial to policydesign. By assigning an intended behavior pattern as a set of training examples for a givenSDM, it can learn a new system structure that fits the data; the differences between theoriginal and new structures lead to considerations of policy design. In short, the neuralrepresentation for SDMs provides a new dimension of studying SD, and some traditionallyhard problems in a FD might now be solved easier in the new representation. It is proved inmathematics that the two model representations operate under the same numericalpropagation constraints.
机译：本文涉及一种用于神经网络的人工神经网络（ANN）表示。 系统动力学模型（SDM）及其在模型构建和策略设计中的优势。 已经注意到SDM和ANN之间的相似之处，即两者都存储 知识主要在于模型的结构（或链接），而不是单元或其他知识 成分。通过专门设计的映射方案，可以看到给定的流程图 （FD）（即SDM的传统表示形式）可以映射到 局部递归网络（PRN）的表示将正确地表现得像一个 它模仿。由于可以使用示例数据来训练（部分递归）神经网络， 可以通过从一组 归纳数据的多元时间序列。这为SD的研究增添了优势，因为 可以通过学习而不是手动构建来创建模型，这完全是 依靠专家的观察和推论质量。同样，对政策也有好处 设计。通过将预期的行为模式分配为给定的一组训练示例 SDM，它可以学习适合数据的新系统结构;之间的差异 原始结构和新结构导致了对政策设计的考虑。简而言之，神经 SDM的表示形式提供了研究SD的新维度，传统上 现在，在新的表示形式中，可以更轻松地解决FD中的难题。事实证明 两个模型表示在相同数值下运行的数学 传播约束。

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