Overview

JBoost is an implementation of boosting in java. The package you download includes the source, the executable java, visualization scipts (mostly written in python) and a collection of examples that demonstrate the capabilities of Jboost.

Some of the algorithms currently implemented include AdaBoost, LogitBoost, RobustBoost, Boostexter and alternating decision trees. For more information on these algorithms see the publications page

The software is designed to be used in batch mode through a command line interface. The reason is that learning can take a long time. However, information on the progress of learning is printed into files as the software executes, so long runs can be monitored while the program is running.

To learn more, download JBoost or read the documentation.

New in Version 2.0!

The following are the new features of JBoost 2.0:

• RobustBoost support added -- a new boosting algorithm that is resistant to label noise.
• A new visualization tool -- the score visualizer
• Support for stopping and restarting the boosting process while eliminating those examples with small weight from the restarted process.
• JBoost no longer supports Multi-class problems internally, but now offers a wrapper script.