论文部分内容阅读
神经元的映射分配是人工神经网络虚拟实现中的重要研究课题.本文系统地分析了人工神经网络的重要性质──并行分布处理,并对映射分配问题中的两个关键性概念──负载均衡和通信开销进行了深入讨论.以此为基础,提出了一系列映射算法,并对算法性能进行了分析.其中,吸收算法最大程度地开发了人工神经网络固有的并行性,是一个实时的算法.
Mapping and distribution of neurons is an important research topic in the artificial realization of artificial neural network. This paper systematically analyzes the important property of artificial neural networks ─ ─ parallel distribution processing, and the two key concepts in the mapping assignment ─ ─ load balancing and communication overhead are discussed in depth. Based on this, a series of mapping algorithms are proposed, and the performance of the algorithm is analyzed. Among them, the absorption algorithm to maximize the inherent development of the artificial neural network parallelism, is a real-time algorithm.