University of Leicester
Browse
draft.pdf (4.03 MB)

Multiview Subspace Clustering via Low-Rank Symmetric Affinity Graph

Download (4.03 MB)
journal contribution
posted on 2023-03-27, 10:21 authored by W Lan, T Yang, Q Chen, S Zhang, Y Dong, Huiyu Zhou, Y Pan

Multiview   subspace   clustering   has   been   used   toexplore the internal structure of multiview datasets by revealingunique information from different views. Most existing methodsignore  the  consistent  information  and  angular  information  ofdifferent  views.  In  this  paper,  we  propose  a  multiview  subspaceclustering method (LSGMC) based on low-rank consistency andsymmetric affinity graph. Specifically, considering the consistentinformation,  we  pursue  a  consistent  low-rank  structure  acrossviews  by  decomposing  the  coefficient  matrix  into  three  factors.Then,  the  symmetry  constraint  is  utilized  to  guarantee  weightconsistency for each pair of data samples. In addition, consideringthe  angular  information,  we  utilize  the  fusion  mechanism  tocapture  the  inherent  structure  of  data.  Further,  to  alleviate  theeffect  brought  by  the  noise  and  the  high  redundant  data,  theSchatten  p-norm  is  employed  to  obtain  a  low-rank  coefficientmatrix.  Finally,  an  adaptive  information  reduction  strategy  isdesigned to generate a high-quality similarity matrix for spectralclustering.  Experimental  results  on  eleven  datasets  demonstratethe  superiority  of  LSGMC  in  clustering  performance  comparedwith ten state-of-the-art multiview clustering methods.

History

Author affiliation

School of Computing and Mathematical Sciences, University of Leicester

Version

  • AM (Accepted Manuscript)

Published in

IEEE transactions on neural networks and learning systems

Publisher

IEEE

issn

2162-2388

Copyright date

2023

Available date

2023-03-27

Language

en

Usage metrics

    University of Leicester Publications

    Categories

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC