复杂系统拓扑结构对人工神经网络的优化综述  

Review of Complex System Topology for Artificial Neural Network Optimization

作　　者：林志诚 马永航 LIN Zhicheng;MA Yonghang(Xinjiang University,Urumqi 830046,China)

机构地区：[1]新疆大学,乌鲁木齐830046

出　　处：《移动信息》2023年第11期129-131,135,共4页MOBILE INFORMATION

摘　　要：随着人工神经网络在不同领域的成功应用,改变网络结构以优化其性能成为近年来的研究热点。由于人工神经网络具有广泛的连通性和复杂的结构,在获得高性能的同时,设计时间、布线成本和空间成本都更低的稀释型人工神经网络备受关注。复杂系统理论主要考虑结构对网络整体行为的影响,并将其应用于人工神经网络,使其具有更高的效率和更简单的结构。研究表明,复杂随机拓扑结构在连通性较低的情况下也要优于全连接人工神经网络。但根据神经生物学的研究,具有短特征路径长度和无标度分布的高度聚集的神经元更受青睐,且连接成本更低。因此,将小世界和无标度拓扑应用于人工神经网络成了相关领域的研究热点。文中总结和讨论了小世界、无标度和混合复杂网络与传统的全连接和随机结构对人工神经网络性能的影响。With the successful application of artificial neural networks in different fields,changing the network structure to optimize its performance has become a research hotspot in recent years.Due to the extensive connectivity and complex structure of artificial neural networks,diluted artificial neural networks with lower design time,wiring cost and space cost have attracted much attention while obtaining high performance.Complex system theory mainly considers the influence of structure on the overall behavior of the network,and applies it to artificial neural networks to make them have higher effi-ciency and simpler structure.Studies have shown that complex random topology is also better than fully connected artifi-cial neural networks under low connectivity.However,according to neurobiological research,highly aggregated nerve cells with short characteristic path lengths and scale-free distributions are preferred,and the connection cost is lower.There-fore,the application of small-world and scale-free topology to artificial neural networks has become a research hotspot in related fields.This paper summarizes and discusses the effects of small-world,scale-free and hybrid complex networks and traditional fully connected and random structures on the performance of artificial neural networks.

关 键 词：拓扑结构 人工神经网络 复杂系统 小世界网络 随机网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]