This repository contains MATLAB code accompanying the paper:
"Feature-level domain adaptation."
which is published in the Journal of Machine Learning Research 2016 (pdf/preprint).
For a cleaner implementation of flda as well as a translation into Python, see my library on transfer learners and domain-adaptive classifiers: libTLDA.
Clone the repository (bash):
git clone https://github.com/wmkouw/fldaInstallation consists of adding the repository to your path (matlab):
addpath(genpath('./flda'))
Flda depends on minFunc and libSVM. First download and extract them (bash):
wget http://www.cs.ubc.ca/~schmidtm/Software/minFunc_2012.zip -O minFunc.zip
unzip minFunc.zip
wget http://www.csie.ntu.edu.tw/~cjlin/cgi-bin/libsvm.cgi?+http://www.csie.ntu.edu.tw/~cjlin/libsvm+zip -O libSVM.zip
unzip libSVM.zip
Then add them to your path (matlab):
addpath(genpath('./minFunc_2012'))
addpath(genpath('./libSVM-3.22'))
Repo contains the following folders:
- experiment-*: contains scripts for running experiments reported in the paper.
- data: contains the digits, spam, office, imdb and amazon data sets.
- util: contains utility functions and algorithms.
To start an experiment, call the corresponding experiment function (matlab):
cd experiment-amazon/
run_daexp_amazon('flda_log_b')
Options for classifiers are:
- 'flda_log_b': flda with logistic loss and blankout transfer model
- 'flda_log_d': flda with logistic loss and dropout transfer model
- 'flda_qd_b': flda with quadratic loss and blankout transfer model
- 'flda_qd_d': flda with quadratic loss and dropout transfer model
- 'gfk_knn': geodesic flow kernel with a k-nearest-neighbour classifier
- 'tca_svm': transfer component analysis with a support vector machine
- 'sa_svm': subspace alignment with a support vector machine
- 'kmm': kernel mean matching with importance-weighted logistic regression
- 'scl': structural correspondence learning with logistic regression
Bugs, comments and questions can be submitted to the issues tracker.