Multi view Subspace Clustering using Good Neighbors

Abstract

We consider the problem of clustering N data points fxigN i=1 2 Rp, into K number of clusters. We are dealing with high dimensional data points in our scenario where p >> N, i.e. the number of features is much greater than the number of data points. In our work, we set out to solve this problem using subspace clustering, assuming that our high dimensional data points lie in an union of low dimensional subspaces. We try to solve the problem of clustering in the context of multi view data. We find the self-expression matrices from each of the views using Entropy Norm formulation. Then, we find the consensus self-expression matrix by taking the average of all the individual self-expression matrices. Finally, we apply good neighbors post processing to obtain a sparser and strongly connected self-expression matrix thus resulting in an improved affinity graph. The resultant clusters are obtained using Normalized spectral clustering.