k-Manifolds: Nonlinear Subspace Clustering

This work extends manifold learning to classify and parameterize data which lie on multiple, intersecting manifolds. Our implementation introduces technical contributions, including node-weighted MDS, which may be of broader interest. We show examples for intersecting manifolds of mixed topology and dimension.

Read our ICCV 2005 paper.

Download the Matlab source code.
This work incorporates some new modifications not described in the ICCV paper and works only for manifolds which are topologically planar.  This code requires the free versions of Isomap and Netlab to operate correctly.

If you have any questions or are going to use this work in your research, please e-mail me to let me know.