计算机技术与发展2025，Vol.35Issue(8)：61-68,8.DOI:10.20165/j.cnki.ISSN1673-629X.2025.0081
基于典型相关性的多视图方法分析及调用策略框架
Research Analysis on Multi-view Methods Based on Canonical Correlation Analysis and Its Calling Strategy Framework
摘要
Abstract
In the era of big data,data comes from various sources or is extracted from different perspectives,which is commonly referred to as multi-view data.Delving into the interrelations among multiple views of data holds significant importance and practical value.The most popular multi-view analysis method currently is based on canonical correlation analysis(CCA).In recent years,numerous canonical multi-view correlation analysis methods have been researched and proposed.These methods primarily determine the best pair of projection vectors by using transformations and Pearson's coefficient to reveal the maximum correlation between data sets.However,the focus has mainly been on comparative optimization studies between methods rather than integrating different CCA multi-view methods for different scenarios.We aim to compare and analyze some representative typical multi-view correlation analysis methods in recent years,and discuss their performance in image classification.Experiments are conducted on six public benchmark datasets,and their accuracy,robustness,and time complexity are compared and analyzed.Consequently,a framework for integrating multiple CCA multi-view methods for different scenarios is proposed,filling the gap in this area and providing a reference for further research on the optimization and application of multi-view CCA methods.关键词
典型相关分析/多视图数据/图像分类/模型调用策略/特征提取Key words
canonical correlation analysis/multi-view data/image processing/model calling strategy/feature extraction分类
信息技术与安全科学引用本文复制引用
钟海舒,梁瑾..基于典型相关性的多视图方法分析及调用策略框架[J].计算机技术与发展,2025,35(8):61-68,8.基金项目
国家重点研发计划-"战略性科技创新合作"重点专项(2023YFE0207800) （2023YFE0207800）
