Code for the ICML 2013 and AAAI 2013 papers on ELLA
Switch branches/tags
Nothing to show
Clone or download
Latest commit 1ce0495 Mar 10, 2017

README

EFFICIENT LIFELONG LEARNING ALGORITHM (ELLA)

Version 1.0  (July 3, 2013)

Copyright (c) 2013  Paul Ruvolo & Eric Eaton

The copyright extends to all implementation files for ELLA under this distribution.

This program is free software; you can redistribute it and/or modify it under the terms 
of the GNU General Public License Version 3 

 

BUILDING

The ELLA code is written in pure matlab and thus does not need to be built.
However, ELLA does have a dependency on the Sparse Additive Modeling Toolbox of
Julien Mairal et al. (http://spams-devel.gforge.inria.fr/).  In order to use
ELLA one must first go to the subdirectory 'externalLibs/spams-matlab' and
follow the instructions for building this toolbox.



RUNNING THE CODE

We have included three demos for utilizing the different aspects of the ELLA
software.  The demo files are called:
	(1) runExperiment
	(2) runExperimentActiveTask
	(3) runExperimentActiveTaskTargeted



CITING OUR WORK

If you use or adapt this code, please cite the following papers:

Paul Ruvolo & Eric Eaton. (2013). ELLA: An efficient lifelong learning algorithm. In Proceedings of the 30th International Conference on Machine Learning.  Atlanta, GA, June.

Paul Ruvolo & Eric Eaton. (2013). Scalable lifelong learning with active task selection. In Proceedings of the 27th AAAI Conference on Artificial Intelligence.  Bellevue, WA, July.



DATA

This release includes the landmine data set for use by the demonstration scripts. 
The landmine data was originally described in:

Xue, Y., Liao, X., Carin, L., & Krishnapuram, B. (2007). Multi-task learning for classification with Dirichlet process priors. The Journal of Machine Learning Research, 8, 35-63.

Please cite this paper if you use the landmine data set in your own work.