Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

CUR Decompositions, Similarity Matrices, and Subspace Clustering

About

A general framework for solving the subspace clustering problem using the CUR decomposition is presented. The CUR decomposition provides a natural way to construct similarity matrices for data that come from a union of unknown subspaces $\mathscr{U}=\underset{i=1}{\overset{M}\bigcup}S_i$. The similarity matrices thus constructed give the exact clustering in the noise-free case. Additionally, this decomposition gives rise to many distinct similarity matrices from a given set of data, which allow enough flexibility to perform accurate clustering of noisy data. We also show that two known methods for subspace clustering can be derived from the CUR decomposition. An algorithm based on the theoretical construction of similarity matrices is presented, and experiments on synthetic and real data are presented to test the method. Additionally, an adaptation of our CUR based similarity matrices is utilized to provide a heuristic algorithm for subspace clustering; this algorithm yields the best overall performance to date for clustering the Hopkins155 motion segmentation dataset.

Akram Aldroubi, Keaton Hamm, Ahmet Bugra Koku, Ali Sekmen• 2017

Related benchmarks

TaskDatasetResultRank
Motion SegmentationHopkins 155 (all sequences)
Mean Clustering Error0.36
57
Motion SegmentationHopkins 155 2-motion sequences
Classification Error0.0017
36
Subspace ClusteringHopkins155 3 motions (Checker (26))
Average Error3.25
10
Subspace ClusteringHopkins 155 2 motions (Checker 78)
Average Classification Error0.94
10
Subspace ClusteringHopkins 155 2 motions (Traffic 31)
Average Classification Error (%)1.08
10
Subspace ClusteringHopkins155 3 motions (all (35 seq))
Average Classification Error3.63
10
Subspace ClusteringHopkins 155 2 motions (All (120 seq))
Average Classification Error1.47
10
Subspace ClusteringHopkins 155 seq (all sequences)
Average Classification Error1.96
10
Subspace ClusteringHopkins155 3 motions (Traffic 7)
Average Classification Error3.57
10
Subspace ClusteringHopkins155 3 motions (Articulated (2))
Average Classification Error8.8
10
Showing 10 of 14 rows

Other info

Follow for update