计算机应用研究2025，Vol.42Issue(10)：2881-2892,12.DOI:10.19734/j.issn.1001-3695.2025.03.0066
胶囊网络综述
Overview of capsule networks
摘要
Abstract
Capsule networks effectively overcome the data dependency of convolutional neural networks and their limitations in modeling object position,pose,and spatial relationships through their unique network structure and dynamic routing algo-rithm,demonstrating significant potential and adaptability.To systematically review the research progress of capsule network's architecture,this article discussed fundamental theories and optimization strategies,experimental comparisons,and application domains.Firstly,it introduced the basic architectures and core algorithms of vector capsule networks and matrix capsule networks,and analyzed various optimization strategies along with their advantages and disadvantages.Next,experimental com-parisons evaluated the performance of capsule networks and their improved models.Furthermore,this paper explored applica-tions of capsule networks in object detection,healthcare,and transportation.Finally,it outlines future research directions and development trends to guide further advancements in the field.关键词
胶囊网络/卷积神经网络/姿态/向量胶囊网络/矩阵胶囊网络/优化策略Key words
capsule network/convolutional neural network/pose/vector capsule network/matrix capsule network/optimiza-tion strategies分类
计算机与自动化引用本文复制引用
朱晓娟,李卫军,马馨瑜,熊章友,杨国梁..胶囊网络综述[J].计算机应用研究,2025,42(10):2881-2892,12.基金项目
宁夏高等学校科学研究项目(NYG2024086) （NYG2024086）
