Please read the README file for copyright information. I do not take any responsibility on damage or other problems caused by using these programs. But you are welcomed to discuss with me if you have difficulty applying them for your needs.

RankBoost and ORBoost algorithms

The program performs ordinal regression using the large-margin thresholded ensemble model with various algorithms. One of them is an efficient implementation of the RankBoost algorithm (Freund et al., 2003) with thresholding. The other two are our proposed ORBoost algorithms. Details can be found in

H.-T. Lin and L. Li. Large-Margin Thresholded Ensembles for Ordinal Regression: Theory and Practice, 2006.

The latest implementation was finished in May 2006, and was later polished in June 2007. You can download the source code here.

Usage Example:

./boostrank-train ~/pyrimdines.dense 50 27 30 200 10 100 pyrimdines.model

• 50 is the number of training examples
• 27 is the number of features
• 30 means ORBoost-All
• 200 means with perceptron (RCD) weak learner
• 10 means number of ranks
• 100 means number of iterations

Note: the file pyrimdines.dense was processed from the files provided by Dr. Wei Chu here.

Feel free to contact me: "htlin" at "csie.ntu.edu.tw"