四川大学学报（自然科学版）2025，Vol.62Issue(3)：751-760,10.DOI:10.19907/j.0490-6756.240143
鲁棒图正则化多视角函数型聚类算法
A robust graph regularized multi-view functional clustering algorithm
摘要
Abstract
Multi-view clustering is an important data analysis and mining tool in machine learning and artifi-cial intelligence.By integrating and utilizing the difference and complementary information between different perspectives,this method can fully reveal the internal structure and characteristics of dataset and give interpre-tive clustering results.However,clustering tasks can be detrimentally damaged by noise and superfluous data in multi-view dataset.In this paper,based on the non-negative matrix factorization method we propose a new functional clustering algorithm,that is,robust graph regularized multi-view functional clustering algorithm(RGMFC).In the algorithm,the structured sparse l2,1-norm is used to suppress the influence of noise or outli-ers,and the graph regularization strategy is introduced and multi-view heterogeneous features are integrated considering the local geometric characteristics of dataset.Simulation experiments implemented on both the random simulated dataset and the Growth dataset demonstrate that the algorithm can enhance the clustering performance and exhibit strong robustness.Finally,the effectiveness of the algorithm is validated through the practical application of identifying the spatial layout of air quality monitoring stations in Beijing.关键词
非负矩阵分解/多视角函数型聚类/鲁棒性/图正则化Key words
Non-negative matrix factorization/Multi-view functional clustering/Robustness/Graph regular-ization分类
信息技术与安全科学引用本文复制引用
高海燕,程莞莞..鲁棒图正则化多视角函数型聚类算法[J].四川大学学报（自然科学版）,2025,62(3):751-760,10.基金项目
国家社会科学基金(19XTJ002) （19XTJ002）
甘肃省自然科学基金(23JRRA1186) （23JRRA1186）
甘肃省高校青年博士支持项目(2025QB-058) （2025QB-058）
