Automatic Moment-Closure Approximation of Spatially Distributed Collective Adaptive Systems
Automatic Moment-Closure Approximation of Spatially Distributed Collective Adaptive Systems
复制标题
空间分布式集体自适应系统的自动矩收敛逼近
DOI：
10.1145/2883608
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Vashti Galpin
中科院分区：
文献类型：
--
作者：
Cheng Feng;J. Hillston;Vashti Galpin
空间分布集体自适应系统是一类重要的系统，由于其状态空间的大小和复杂性，对建模提出了重大挑战。当必须捕获系统的动态行为时，例如为了预测系统性能，这个问题是尖锐的。在这篇文章中，我们提出了一种抽象技术，自动推导出一个时刻封闭近似的动态行为的空间分布的集体自适应系统从离散表示的实体。力矩闭合技术被证明可以准确估计动态行为，虽然在某些情况下，二阶联合力矩产生的常微分方程的数量可以增长很大。对于这些情况下，我们提出了一个严格的模型降阶技术，并证明了它的使用，大大减少了计算工作量，只有有限的影响，如果减少阈值设置适当的准确性。本文中报告的所有技术都在一个可免费下载的工具中实现。
Spatially distributed collective adaptive systems are an important class of systems that pose significant challenges to modeling due to the size and complexity of their state spaces. This problem is acute when the dynamic behavior of the system must be captured, such as to predict system performance. In this article, we present an abstraction technique that automatically derives a moment-closure approximation of the dynamic behavior of a spatially distributed collective adaptive system from a discrete representation of the entities involved. The moment-closure technique is demonstrated to give accurate estimates of dynamic behavior, although the number of ordinary differential equations generated for the second-order joint moments can grow large in some cases. For these cases, we propose a rigorous model reduction technique and demonstrate its use to substantially reduce the computational effort with only limited impact on the accuracy if the reduction threshold is set appropriately. All techniques reported in this article are implemented in a tool that is freely available for download.
DOI：
10.1016/j.tcs.2010.02.001
发表时间：
2010-05
期刊：
Theor. Comput. Sci.
影响因子：
--
作者：
R. A. Hayden;J. Bradley
通讯作者：
R. A. Hayden;J. Bradley