Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Latest commit 77a27a3 Oct 16, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
SSC_ADMM_v1.1 UOPC Oct 16, 2019
Cone_Algo_Compare.m UOPC Oct 16, 2019
Cone_Algo_Compare_Motion.m UOPC Oct 16, 2019
KNNG_M.m UOPC Oct 16, 2019
KNNG_UOPCA.m UOPC Oct 16, 2019
MNIST.mat UOPC Oct 16, 2019
MotionData1.mat UOPC Oct 16, 2019
MotionData2.mat UOPC Oct 16, 2019
NCL_UOPC.m UOPC Oct 16, 2019
PureKNN.m UOPC Oct 16, 2019
README.md UOPC Oct 16, 2019
Readme.rtf UOPC Oct 16, 2019
Readme.txt UOPC Oct 16, 2019
TSC_Update.m UOPC Oct 16, 2019
main_2D.m UOPC Oct 16, 2019
main_3D.m UOPC Oct 16, 2019
main_MNIST.m UOPC Oct 16, 2019
main_Motion.m UOPC Oct 16, 2019
main_motion_TSC.m UOPC Oct 16, 2019
parsave.m UOPC Oct 16, 2019

README.md

UOPC

Copyright @ Wenqi Wang, 2017

The proposed KNN-SC algorithm is KNNG_UOPCA.m in the document. To replicate the reults, cvx is needed to be downloaded from (http://cvxr.com/cvx/) with Mosek (http://cvxr.com/cvx/doc/mosek.html) solver installed.

To run the code

(1) run main_2D.m for 2 dimensional synthetic data

(2) run main_3D.m for 3 dimensional synthetic data

(3) run main_MNIST for MNIST data clustering

(4) run main_Motion for motion data clustering

Note that the comparables in the paper are

(1) Sparse Subspace Clustering (SSC) (http://vision.jhu.edu/code/)

(2) robust subspace clustering via thresholding (TSC) (https://www.nari.ee.ethz.ch/commth/research/downloads/sc.html)

(3) Non-negative Contraint Lasso (NCL), which is implemented using Mosek solver in cvx (http://cvxr.com/cvx/)

When use the code, please cite the following paper:

Unsupervised clustering under the Union of Polyhedral Cones (UOPC) model, Wenqi Wang, Vaneet Aggarwal, and Shuchin Aeron, Pattern Recognitino Letters, Vol 100, pp 104-109 Available: https://www.sciencedirect.com/science/article/pii/S0167865517303653

You can’t perform that action at this time.