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.
