splitsvm v.0.9 
~~~~~~~~~~~~~~

Dependancies:
`````````````
trove-2.0.2.jar  (included in the lib/ directory)


compiling:
`````````````
> ant jar

This will produce splitsvm.jar


Using:
`````````````
Note that splitsvm is used only for SVM *prediction*.  
For training, you need an external SVM package.
Currently, YAMCHA and LIBSVM models are supported.

Once you have a LIBSVM / YAMCHA model, you need to convert it into a splitsvm
model.  This is done by a supplied python script (required python2.5 in order to
run).

libsvm usage
``````````````
assuming you have a libsvm model called 'abc.mod'

(1) convert it to splitsvm format:
> python/mksplitsvm.py libsvm abc.mod abc 

this will create a file called abc.spltmod

(you can run the mksplitsvm.py script without any parameters to get its usage instructions)

If desired (though not neccesary) you can then convert the model to a binary
form, so that it loads faster (about twice as fast):
> java -cp splitsvm.jar yg.classifier.splitsvm.io.MakeBinaryModel abc.spltmod


(2) change your code to use splitsvm instead of libsvm for classification
   
libsvm code:
"""
import libsvm.svm;
import libsvm.svm_model;
import libsvm.svm_node;

public class LIBSVMExample {
   public static void main(String[] args) {
      
      svm_model m = svm.svm_load_model("abc.mod");
      svm_node[] x = ... // whatever

      int p = svm.svm_predict(m, x);
   }
   
}
"""

splitsvm code:
"""
import yg.classifier.splitsvm.FastMulticlassModel;
import yg.classifier.splitsvm.wrappers.FastLibSVM;

import libsvm.svm_node;

public class SplitSVMExample {
   public static void main(String[] args) {
      
      FastMulticlassModel m = svm.svm_load_model("abc.spltmod");
      svm_node[] x = ... // whatever

      int p = FastLibSVM.svm_predict(m, x);
   }
   
}
"""