A library for large linear classification, forked from

http://www.csie.ntu.edu.tw/~cjlin/liblinear/

I needed to make source-level modifications. In particular, I wanted to optimize parameters using measures other than accuracy. The current version uses F-Score: 2 * Precision * Recall / (Precision + Recall).

1. The problem described by the input must be a binary classification problem, and class labels must be either +1 or -1.

2. Modify the assigned value of validation_function in eval.cpp to change the evaluation function.

3. Note the current Makefile is tweaked for OS X. Use Makefile.orig if you're on another unix-style OS.

4. See README for more details.