计算机科学与探索 ›› 2021, Vol. 15 ›› Issue (2): 284-293.DOI: 10.3778/j.issn.1673-9418.1912033

• 人工智能 • 上一篇    下一篇

基于核诱导的不完整多视角聚类

张炜,邓赵红,王士同   

  1. 江南大学 人工智能与计算机学院,江苏 无锡 214122
  • 出版日期:2021-02-01 发布日期:2021-02-01

Kernel-Induced Incomplete Multi-view Clustering

ZHANG Wei, DENG Zhaohong, WANG Shitong   

  1. School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2021-02-01 Published:2021-02-01

摘要:

随着技术的发展,数据往往具有来自不同源的多种形式,多视角聚类算法旨在利用不同源中的互补信息进行聚类。虽然目前多视角聚类算法已在各个领域取得较大发展和成功应用,但是多视角聚类算法仍然面临许多重要挑战,其中一个就是当多个视角的样本存在缺失时,如何充分挖掘数据信息以减少缺失样本带来的负面影响。针对此挑战,提出一种基于核诱导的不完整多视角聚类算法(KIMV)。该方法利用核方法和非负矩阵分解技术在核希尔伯特空间中对所有视角学习一个最优的共性矩阵,并通过视角自适应加权机制和图拉普拉斯正则化提高算法性能。在五个多视角数据集上的实验有效验证了KIMV的上述优势。

关键词: 不完整多视角, 核诱导, 拉普拉斯正则化, 自适应视角加权

Abstract:

With the development of technology, data often have multiple forms which come from multiple sources. The multi-view clustering algorithm aims to use the complementary information existing in different sources for clustering. Although the multi-view clustering algorithm has successful applications in many fields, it still faces many challenges, one of which is how to fully mine information between views to reduce the negative impact of missing instances when multi-view data are incomplete. To this end, the algorithm kernel-induced incomplete multi-view clustering (KIMV) is proposed in this paper. On one hand, the kernel method and the non-negative matrix factorization are used to learn an optimal latent feature matrix in the kernel Hilbert space, on the other hand, the performance of the algorithm is improved by adaptive weighting mechanism and Laplace regularization. Experiments on five datasets demonstrate the advantages of KIMV.

Key words: incomplete multi-view, kernel-induced, Laplacian regularization, adaptive view weighting