Learning algorithm described in "A New PAC-Bayesian Perspective on Domain Adaptation" (see http://arxiv.org/abs/1506.04573)
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore
LICENSE
README
__init__.py
common.py
dalc.py
dalc_classify.py
dalc_learn.py
dalc_reverse_cv.py
dataset.py
kernel.py

README

---------------------------------------------------------------------------
DOMAIN ADAPTATION OF LINEAR CLASSIFIERS (aka DALC)
Version 0.90 (November 2, 2015), Released under the BSD-license
---------------------------------------------------------------------------
Author:
    Pascal Germain. Groupe de Recherche en Apprentissage Automatique
    de l'Universite Laval (GRAAL).

Reference:
    Pascal Germain, Amaury Habrard, Francois Laviolette, and Emilie Morvant.
    A New PAC-Bayesian Perspective on Domain Adaptation.
    International Conference on Machine Learning (ICML) 2016.
    http://arxiv.org/abs/1506.04573
----------------------------------------------------------------------------

Thank you for looking at my code!

This program have been tested using Python 3.6 under Linux and MacOS.
It requires the NumPy and SciPy libraries.

I prepared three small scripts to use DALC by the command line:
1) dalc_learn.py: Execute the learning algorithm
2) dalc_classify.py: Execute the classification function
3) dalc_reverse_cv.py: Compute a "reverse cross-validation" score

Further usage instructions can be obtained by the following commands:
python dalc_learn.py --help
python dalc_classify.py --help
python dalc_reverse_cv.py --help

The data used in the paper experiments is available here (in svmlight format):
http://researchers.lille.inria.fr/pgermain/data/amazon_tfidf_svmlight.tgz

Pascal Germain.