K-subspaces

K-subspaces provides the k-means like clustering.

The K-subspaces provides the k-means like clustering.
As same as the k-means clustering, subspaces are randomly initialized.
Then, data are classified based on the distance to the subspaces. Actually, the inner products are evaluated instead of the distance. The subspaces are updated based on the PCA with the clustered data.
These processes are iteratively applied to obtained the subspaces and clustering results.