I am glad to present the Learning Classifier System Open Repository (LCSOR); a repository on LCSs implemented in C++ to further extend these family of machine learning techniques. The files are stored in SourceForge.
The project is divided in three distinct parts: (1) the source code (a tarball for each algorithm), (2) the problem environment, and (3) the documentation in the form of doxygen html.
To start I uploaded the first version of the sUpervised Classifier System (UCS). The current version of this algorithm is 0.96, which implements the real interval-based representation (more precisely the unordered-bound representation) and the categorial representation of UCS. Also, the imbalanced estimators are included (see (Orriols-Puig, 2008)). I have codified some other measures of quality: Cohen's Kappa score, precision, recall and the F-Measure (F1).
As you see, this project is just in its infancy and will grow little by little.
Algorithms included so far:
This new version (5/23/2014) incorporates:
- Cohen's Kappa score.
- Minor bug fixes.
Installation detailsIn order to compile and install this software follow these steps:
- download the source code (e.g., UCS.tar.gz),
- uncompress the file using tar ($ tar xvzf UCS.tar.gz),
- enter to the directory containing the source code and compile it ($ make),
- download the problem environment (e.g., Problems.tar.bz),
- uncompress the file using tar ($ tar xvjf Problems.tar.bz), and
- execute the program (e.g, $ ./UCS ../Problems/tao/config/tao.cfg).
If you use LCSOR for your research, please cite it with the following:
author = "Sancho-Asensio, Andreu",
year = "2014",
title = "Learning classifier system open repository",
url = "http://andreusancho.blogspot.com.es/p/lcsor.html",
institution = "Ramon Llull University"
If you have any comments or find any bug, please send a message!