mashaan14 / ASC-self-tuned-k

ASC-self-tuned-k

Approximate spectral clustering with eigenvector selection and self-tuned k

This is an implementation for the following paper:

@article{ALSHAMMARI201931,
	title =   {Approximate spectral clustering with eigenvector selection and self-tuned k},
	doi =     {https://doi.org/10.1016/j.patrec.2019.02.006},
	journal = {Pattern Recognition Letters},
	volume =  {122},
	pages =   {31-37},
	year =    {2019},	
	author =  {Mashaan Alshammari and Masahiro Takatsuka}
}

How to use:

Run BATCH_Points.m which will execute the following:

1. PRE_Points.m to load toy data, csv files are the groundtruth labels.
2. RUN_Points.m to perform spectral clustering with 4 functions to estimate k:
   • CostEigenGap.m a conventional method to estimate k
   • CostZelnik.m uses the method proposed by (Zelnik-manor 2005) to estimate k
   • CostDBIOverLambda.m uses the method proposed by our paper to estimate k
   • CostDBIOverLambdaPCA.m a uses the method proposed by our paper to estimate k followed by PCA variance filtering
3. POST_Points.m to compute the accuracy of clustering