摘要
图卷积神经网络是图论与深度学习的交叉,已成为机器学习领域的研究热点。基于此,介绍了图卷积神经网络的形成,梳理了两大类经典的图卷积神经网络:谱方法和空间方法,详细介绍了这两类图卷积神经网络模型,分析了图卷积操作的核心理论基础,介绍了图卷积神经网络在各领域的应用,总结了图卷积神经网络面临的主要挑战,展望了图卷积神经网络的发展趋势,并分析了图卷积神经网络在野外环境下蝴蝶识别任务中的潜在应用。
Graph convolutional neural network(GCN)has emerged as the intersection of graph theory and deep learning,becoming the hotspot research field of machine learning.Therefore,a comprehensive overview of the GCN is provided,and the available studies of GCN into two typical categories are summarized:spectral-based methods and spatial-based methods.These two typical types of GCN models are extensively discussed,the fundamental theoretical underpinnings of the graph convolution operations are delved into,diverse applications of GCN across various domains are showcased,the major challenges encountered by GCN are summarized,and valuable insights into the future trends of GCN advancement are offerred.Additionally,the potential utilization of GCN in butterfly recognition tasks is investigated,particularly in identifying butterfly species by leveraging images captured in natural habitats.
作者
谢娟英
张建宇
XIE Juanying;ZHANG Jianyu(School of Computer Science,Shaanxi Normal University,Xi'an 710119,Shaanxi,China)
出处
《陕西师范大学学报(自然科学版)》
CAS
CSCD
北大核心
2024年第2期89-101,共13页
Journal of Shaanxi Normal University:Natural Science Edition
基金
国家自然科学基金(62076159,61673251,12031010)
中央高校基本科研业务费项目(GK202105003)。
关键词
图卷积神经网络
谱方法
空间方法
目标检测
graph convolutional neural network
spectral methods
spatial methods
object detection