通信学报 ›› 2022, Vol. 43 ›› Issue (7): 143-152.doi: 10.11959/j.issn.1000-436x.2022134

• 学术论文 • 上一篇    下一篇

面向视角非对齐数据的多视角聚类方法

李骜1, 冯聪1, 牛宇童1, 徐士彪2, 张英涛3, 孙广路1   

  1. 1 哈尔滨理工大学计算机科学与技术学院,黑龙江 哈尔滨 150080
    2 北京邮电大学人工智能学院,北京 100876
    3 哈尔滨工业大学计算机科学与技术学院,黑龙江 哈尔滨 150001
  • 修回日期:2022-06-21 出版日期:2022-07-25 发布日期:2022-06-01
  • 作者简介:李骜(1986- ),男,黑龙江哈尔滨人,博士,哈尔滨理工大学副教授、博士生导师,主要研究方向为模式识别、机器学习等
    冯聪(1997- ),男,广东佛山人,哈尔滨理工大学硕士生,主要研究方向为机器学习、数据挖掘等
    牛宇童(1997- ),男,黑龙江哈尔滨人,哈尔滨理工大学硕士生,主要研究方向为机器学习、视觉目标检测等
    徐士彪(1985- ),男,吉林舒兰人,博士,北京邮电大学教授,主要研究方向为三维计算机视觉、模式识别、深度学习、遥感图像处理及虚拟现实等
    张英涛(1979- ),女,黑龙江哈尔滨人,博士,哈尔滨工业大学副教授,主要研究方向为人工智能与信息处理等
    孙广路(1979- ),男,黑龙江哈尔滨人,博士,哈尔滨理工大学教授、博士生导师,主要研究方向为机器学习、网络安全等
  • 基金资助:
    国家自然科学基金资助项目(62071157);黑龙江省青年创新人才计划(UNPYSCT-2018203);黑龙江省自然科学基金优秀青年基金(YQ2019F011);黑龙江省高等学校基本科研业务专项资金资助项目(LGYC2018JQ013);黑龙江省博士后启动基金资助项目(LBH-Q19112)

Multiview clustering method for view-unaligned data

Ao LI1, Cong FENG1, Yutong NIU1, Shibiao XU2, Yingtao ZHANG3, Guanglu SUN1   

  1. 1 School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080, China
    2 Artificial Intelligence School, Beijing University of Posts and Telecommunications, Beijing 100876, China
    3 School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China
  • Revised:2022-06-21 Online:2022-07-25 Published:2022-06-01
  • Supported by:
    The National Natural Science Foundation of China(62071157);The University Nursing Program for Young Scholars with Creative Talents in Heilongjiang Province(UNPYSCT-2018203);Heilongjiang Natural Science Foundation for Excellent Young Scholars(YQ2019F011);The Fundamental Research Foundation for University of Heilongjiang Province(LGYC2018JQ013);Postdoctoral Foundation of Heilongjiang Province(LBH-Q19112)

摘要:

如何在视角对齐关系错位时有效进行非对齐多视角学习是一类新的挑战性问题。针对这一问题,提出面向视角非对齐数据的多视角聚类方法。一方面,为了捕获多视角异构特征的跨视角相似度信息,基于多视角非负矩阵分解进行表示学习,将原始特征嵌入一个可度量的低维同构空间。另一方面,在低维同构空间中,以二部图最优匹配模型建模视角对齐关系,并提出参考视角概念将模型推广至多视角情形。将表示学习和视角对齐关系学习整合到统一的Bi-level优化框架,使其在迭代中相互促进,进一步提高模型对视角非对齐数据的学习能力。在视角非对齐数据聚类应用上的大量实验结果表明,相比于8种先进的多视角聚类方法,所提方法在3个数据集上的多项性能指标均取得了较优的性能。

关键词: 聚类分析, 多视角学习, 视角非对齐数据, 非负矩阵分解

Abstract:

A new challenge for multi-view learning was posed by corrupted view-correspondences.To address this issue, an effective multi-view learning method for view-unaligned data was proposed.First,to capture cross-view latent affinity in multi-view heterogenous feature spaces,representation learning was employed based on multi-view non-negative matrix factorization to embed original features into a measurable low-dimensional subspace.Second, view-alignment relationships were modeled as optimal matching of a bipartite graph, which could be generalized to multiple-views situations via the proposed concept reference view.Representation learning and data alignment were further integrated into a unified Bi-level optimization framework to mutually boost the two learning processes, effectively enhancing the ability to learn from view-unaligned data.Extensive experimental results of view-unaligned clustering on three public datasets demonstrate that the proposed method outperforms eight advanced multiview clustering methods on multiple evaluation metrics.

Key words: clustering analysis, multiview learning, view-unaligned data, non-negative matrix factorization

中图分类号: 

No Suggested Reading articles found!