readme

svm的实现源码

BSVM:

*****************************************************************

           COPYRIGHT NOTIFICATION

BSVM can be freely used for research purpose.
Use for commercial purposes is expressly prohibited
without contacting the authors.
BSVM is provided "as is" without express or implied warranty.

Chih-Wei Hsu and Chih-Jen Lin
Department of Computer Science and
Information Engineering
National Taiwan University 
Taipei 106, Taiwan
e-mail: [e-mail]

*****************************************************************

           INSTRUCTIONS
          
1. Create the bsvm directory structure with

        unzip bsvm.zip

   This produces the directory bsvm and several subdirectories.

2. On Unix systems, type `make' to build the `bsvm-train' and `bsvm-predict'
   programs. On other systems, consult `Makefile' to build them or use the
   pre-built binaries (Windows binaries are in the subdirectory `Windows').

3. Find options of bsvm
	
	bsvm-train 

4. Run the sample problems by executing bsvm. A model 
   file is generated

	bsvm-train -t 2 -c 1000  vehicle.scale vehicle_model

5. Test the classifier

	bsvm-predict vehicle.scale vehicle_model classified_result
 
   test data: vehicle.scale (a 4-class classification problem)
   support vectors: vehicle_model
   classified_result: results after classification 

*****************************************************************

           FILE FORMATS and PARAMETERS

The format of training and testing data file is:

<label> <index1>:<value1> <index2>:<value2> ...
.
.
.

<label> is the target value of the training data. For classification,
it should be an integer which identifies a class (multi-class classification
is supported). For regression, it's any real number. <index> is an integer starting from 1, <value> is
a real number. The labels in the testing data file are only used to
calculate accuracy. If they are unknown, just fill this column with a number.

The new release of BSVM borrows the structure of libsvm
(http://www.csie.ntu.edu.tw/~cjlin/libsvm). Similar options are also adopted.

Type bsvm-train to know the usage of bsvm:

For more explanation of these options, please check the README of libsvm.
The main difference from libsvm is -s and -q.
  -s 0: bound-constrained multi-class support vector classification (SVC)
  -s 1: multi-class SVC from solving a bound-constrained problem
  -s 2: multi-class SVC from Crammer and Singer
  -s 3: bound-constrained support vector regression
The option -q gives the size of sub-problems for -s 0, 1, and 3.
The size of sub-problems using -s 2 is always the number of classes.
Note that -v now stands for cross validataion. In BSVM 1.1, -v is
the verbosity.

*****************************************************************

           ADDITIONAL INFORMATION

Chih-Wei Hsu and Chih-Jen Lin
1. A simple decomposition method for support vector machines
http://www.csie.ntu.edu.tw/~cjlin/papers/decomp.ps.gz
2. A comparison on methods for multi-class support vector machines.
http://www.csie.ntu.edu.tw/~cjlin/papers/multisvm.ps.gz
3. Decomposition Methods for Linear Support Vector Machines
http://www.csie.ntu.edu.tw/~cjlin/papers/linear.pdf

Acknowledgments:
The authors thank Chih-Chung Chang
for many helpful disussions and comments.
Part of the software implementation
also benifited from his help.


BSVM uses the opmization solver TRON
by Chih-Jen Lin and Jorge More'.
Please read the COPYRIGHT NOTIFICATION
before using it.
The TRON homepage is in 
http://www.mcs.anl.gov/~more/tron