Auto-weighted multi-view co-clustering via fast matrix factorization

Highlights：

• According to the importance of each view for the clustering task, the proposed approaches automatically learn the weight factor in a re-weighted manner. Moreover, the proposed methods are free parameter which make them be more practical.

• Comparing with graph-based multi-view clustering algorithms, the computational complexity of the proposed methods is same as the traditional K-means algorithm due to the fact that they do not need eigenvalue decomposition in solving process which is heavy computation burden for multi-view data sets.

摘要

•Distinguishing the existing multi-view clustering methods, the proposed approaches involve constraints of indicator matrix in matrix decomposition. Due to the existence of constraints, we can directly acquire the clustering results of samples and features. Thus, the proposed methods are highly efficient for the clustering problem of solving multi-view data sets.•According to the importance of each view for the clustering task, the proposed approaches automatically learn the weight factor in a re-weighted manner. Moreover, the proposed methods are free parameter which make them be more practical.•Comparing with graph-based multi-view clustering algorithms, the computational complexity of the proposed methods is same as the traditional K-means algorithm due to the fact that they do not need eigenvalue decomposition in solving process which is heavy computation burden for multi-view data sets.