Sparse Co-Clustering via Multi-view Matrix Decomposition
Jinbo Bi, Jiangwen Sun, and Tingyang Xu
Department of Computer Science and Engineering, University of Connecticut, Storrs, CT, USA
When multiple views of data are available for
a set of subjects, co-clustering aims to identify
subject clusters that agree across the different
views. We explore the problem of coclustering when the underlying clusters exist in
different subspaces of each view. We propose a multi-view singular value decomposition approach
that simultaneously decomposes multiple data matrices into sparse singular vectors.
This approach is able to group subjects consistently across the views and simultaneously identify
the subset of features in each view that specifies the clusters. Extension to multi-view eigen-decomposition
based on the same idea is also discussed. A simulation study validates that the proposed
algorithm can identify the hypothesized clusters and their associated features. Comparison
with several latest multi-view co-clustering methods on benchmark datasets demonstrates the
superior performance of the proposed approach.
- This method (or a variant of the method) has been used in a bioinformatics problem that identifies subtypes of a complex disease, and the related work has been published at the conference proceedings of BIBM 2013 (here) with its extended version published at BMC Genetics.
- A machine learning paper is in preparation that will characterize the full set of multi-view coclustering methods that are based on matrix decomposition, including singular value decomposition and eigen-decomposition, and will come out soon.
- Click here for the Matlab package that has been implemented and described in our paper.
This is an open source program for non-commercial use only. Please contact either Dr. Jinbo Bi (firstname.lastname@example.org) or Tingyang Xu (email@example.com)
for on-going progress.
Contact Jinbo Bi (firstname.lastname@example.org) for information about this page.