Multiple kernel clustering based on centered kernel alignment
作者：
Highlights：
• We explore a new way to construct MKC methods, viz. kernel-evaluation-based MKC.
• A MKC method based on centered kernel alignment (CKA) is proposed.
• CKA unifies the tasks of clustering and MKL into an optimization problem.
• A two-step iterative algorithm is developed to solve the problem efficiently.
• Clustering experiments on UCI and face datasets show the effectiveness of our method.
摘要
Highlights•We explore a new way to construct MKC methods, viz. kernel-evaluation-based MKC.•A MKC method based on centered kernel alignment (CKA) is proposed.•CKA unifies the tasks of clustering and MKL into an optimization problem.•A two-step iterative algorithm is developed to solve the problem efficiently.•Clustering experiments on UCI and face datasets show the effectiveness of our method.
论文关键词：Clustering,Data fusion,Multiple kernel learning,Centered kernel alignment
论文评审过程：Received 15 May 2013, Revised 31 March 2014, Accepted 5 May 2014, Available online 20 May 2014.
论文官网地址：https://doi.org/10.1016/j.patcog.2014.05.005
