Local linear neighbor reconstruction for multi-view data.pdfVIP

Pattern Recognition Letters 84 (2016) 56–62 Contents lists available at ScienceDirect Pattern Recognition Letters journal homepage: /locate/patrec Local linear neighbor reconstruction for multi-view dataR Linlin Zong a,b, Xianchao Zhang b,?, Hong Yu b, Qianli Zhao a,b, Feng Ding b a Faculty of Electronic Information and Electrical Engineering, Dalian University of Technology, Dalian, China b School of Software, Dalian University of Technology, Dalian, China article info Article history: Received 18 July 2015 Available online 23 August 2016 Keywords: Multi-view similarity Similarity construction Local linear neighbor abstract Graph based multi-view data analysis has become a hot topic in the past decade, and multi-view similarity matrix is fundamental for such tasks. Existing multi-view similarity matrix construction methods cannot learn local geometrical information in the original data space from multiple views simultaneously. Considering the fact that an appropriate similarity matrix is block-wise with intra-class similarity, it is more reasonable to learn a similarity matrix by using local geometrical information in multiple original data space. In this paper, we propose to construct a uni?ed similarity matrix by using local linear neighbors in multiple views. In each view, the similarity matrix can be reconstructed with the weights of the neighbors of each data point in the original space. In multiple views, we seek for a uni?ed similarity matrix which consists of the similarity matrix in each view. The uni?ed similarity matrix can be used for spectral clustering, label propagation and other graph based learning algorithms. Experimental results show that spectral clustering and label propagation algorithms using the uni?ed similarity matrix outperform those using other multi-view similarity matrices, they also outperform typical multi-view spectral clustering algorithms and typical multi-view label propagation algorithms. ? 2016 Elsevier B.V. All rights reserved