This package contains the experiment Matlab codes and datasets used in our Pattern Recognition Letters. All the experiments were carried out using Matlab2012b with an Intel(R) Xeon(R) 8-core CPU computer. If you utilize the codes and / or datasets, please give a citation to the following article:

He Zhang, Zhirong Yang and Erkki Oja.:Improving Cluster Analysis By Co-initializations. Pattern Recognition Letters, 2014. Accepted. To appear.

There are 19 datasets used in our experiments: the folder "datasets_multivariate" contains the datasets with multivariate form and the folder "datasets_graph" contains the corresponding datasets with graph form.  

The main script to run the experiments is "clustering_test.m".

You may need to re-compile the C codes before running "clustering_test.m" by

mex -largeArrayDims sp_factor.c
mex -largeArrayDims sp_factor_ratio.c

The Normalized Cut (a), 1-Spectral codes (b), and Cluster Ensemble Toolbox (c) that were used in the experiments are included for the repetition purpose. If you want their updated implementations, please visit correspondingly:  

(a) http://www.timotheecour.com/software/ncut/ncut.html
(b) http://www.ml.uni-saarland.de/code/oneSpectralClustering/oneSpectralClustering.html
(c) http://www.jstatsoft.org/v36/i09


He Zhang, Project Researcher / PhD candidate

Information and Computer Science Department,
Aalto University School of Science, Espoo, Finland

