This directory includes sources used in the following paper:


Cheng-Hao Tsai, Chieh-Yen Lin and Chih-Jen Lin, Incremental and Decremental Training for Linear Classification, 2014.
You will be able to regenerate experiment results in the paper. However,
results may be slightly different due to the randomness, the CPU speed, 
and the load of your computer.

This code has been tested under 64-bit Linux environments with MATLAB 2009.

Please cite the above article if you find this tool useful. Please also read the COPYRIGHT before using this tool.



System Requirement
==================

This experiment is supposed to be run on UNIX machines. The following
commands are required:
- UNIX commands (mv, ln, etc) 
- bash
- g++
- wget
- make
- bunzip2
- perl
- python2.6 or newer versions.
- MATLAB 2009 or newer versions.
About 1G disk space is required.


Run All Experiments
===================

All things will be automatically done by typing

% ./run.sh

in this directory. For other details, please read the following sections.



Build Programs
==============

Please use the following three lines to compile experiment codes

% make -C liblinear-1.92-inc
% make -C liblinear-1.92-dec



Prepare Data Sets
=================

Please change directory to data/, and type

% ./generate_data.sh

All data sets except CTR data will be automatically generated and preprocessed.



Run Methods
===========

Please change directory to exp-inc/ and exp-dec, respectively, and type

% ln -s ../datasets datasets; ./run.py; ./plot.pl 1; ./plot.pl 1024;

All figures and tables will automatically be generated. Figures will be put in 
directories `exp-inc/photo_C1_eps/', `exp-inc/photo_C1024_eps/', `esp-dec/photo_C1_eps/' 
and `exp-dec/photo_C1024_eps/'. Note that some figures not included in the 
paper will also be generated as supplementary.
