文献类型：期刊文献

中文题名：广义细胞自动机的结构及其硬件实现

英文题名：The Architecture and Circuital Implementation Scheme of A New Generalized Cellular Automata

作者：帅典勋[1,2];冯翔[1,2];赵宏彬[1,2];王兴[1,2]

机构：[1]华东理工大学计算机科学与工程系,上海200237;[2]清华大学智能技术与系统国家重点实验室,北京100080

年份：2004

卷号：27

期号：11

起止页码：1441

中文期刊名：计算机学报

外文期刊名：Chinese Journal of Computers

收录：CSTPCD;;Scopus;北大核心:【北大核心2000】；CSCD:【CSCD2011_2012】；

基金：国家自然科学基金重点项目 (60 13 5 0 10 );国家"九七三"重点基础研究发展规划项目 (G19990 3 2 70 7);国家自然科学基金项目 (60 473 0 44;60 0 73 0 0 8);清华大学智能技术和系统国家重点实验室开放课题基金资助 .

语种：中文

中文关键词：并行结构;并行计算;电路实现;细胞神经网络;广义细胞自动机

外文关键词：parallel architecture; parallel computation; circuital implementation; cellular neural network; generalized cellular automata

摘要：该文作者曾提出了广义细胞自动机 (GCA)的原理和并行算法 ,并且应用于网络快速包交换等动态优化问题 .该文进一步讨论了这种新的广义细胞自动机的体系结构、算法的硬件实现及其电路设计 ,它们对于GCA的实际应用有重要意义 .GCA结构不同于Hopfield神经网络 (HNN)和细胞神经网络 (CNN) ,GCA由多层次多粒度宏细胞组成塔形结构 ,它具有多粒度的宏细胞动力学特征 ,相同粒度宏细胞之间没有交互 ,但不同粒度宏细胞之间存在一定程度的交互或反馈 .分析和实验表明 ,在问题求解的优化性、实时性、硬件实现复杂性等方面 ,该文给出的GCA结构和硬件实现 ,与HNN和CNN相比有诸多优点 .
Authors have proposed the conception and parallel algorithm of a generalized cellular automata (GCA) for effectively solving a class of optimization problems in computer networks, such as the Fast Packet Switching problem. This paper further discusses the issues about the architecture,hardware implementation and circuital design scheme of the GCA, which is essential to the applications of GCA approach. Unlike the Hopfield-type neural network (HNN) and cellular neural network (CNN), the proposed GCA has a pyramid architecture that is composed of multi-layer multi-granularity macro-cells, and has the multi-granularity evolution dynamics. In a GCA there is no direct interconnection among the macro-cells with the same granularity, whereas there are some interactions among different macro-cell layers, which not only significantly improves the real-time performance for problem-solving, but also greatly simplifies the hardware structure of GCA. The GCA architecture and its circuital implementation scheme have advantages over the HNN and CNN methods in terms of the real-time performance, interconnection complexity, and parameter selection.