/*      spkmeans-- A Clustering and Dimension Reduction toolkit

        Copyright(c) 2001 Yuqiang Guan 
*/


This is the README file for spkmeans.

-i f gives the initial seeding file; otherwise the initial clusters will be random-perturb
-c   gives the number of clusters
-K   gives the dimension for dimension reduction
-O   gives name of the output cluster matrix
-t   gives scaling, say 'txx', 'tfn', etc.
-D [c|q|b] does concept decomposition. 
	'c' is concept decomposition, 
	'q' is QR decomposition of concept matrix
	'b' is 'c'+'q'
-T  gives true label file
-e  gives epsilon
-n  disables dump infomation(confusion matrix, purity, etc.)

* If you just want to cluster:

spkmeans -c 3 [-i f initial_seeds_file] [-T true label file] -t tfn -O classic3 ./classic3

* If you just want dimension reduction:

spkmeans -K 10 -i [g|f] [graph-part-file|initial_seeds_file] -t tfn -O classic3 -D [c|q|b] ./classic3

* If you want dimension reduction first and then cluster the dimension-reduced matrix:

spkmeans -c 3 -K 10 -i [g|f] [graph-part-file|initial_seeds_file -t tfn -O classic3 -D [c|q|b] ./classic3


The file name is the prefix of 5 matrix file names (_dim, _col_ccs, _row_ccs, _tfn_nz and _docs)
   
* True label file has the format of

#of Docs. 1
ClusterID [x /path/file]
...

* Initial seeding file has the format of

#of Docs. 1
ClusterID [x /path/file]
...

where ClusterID should cover 0 to #ofCLusters. IDs that are outof this range will be ignored.