Awesome

libTLDA

Library of transfer learners and domain-adaptive classifiers.

This package contains the following classifiers:

• Importance-weighting with ratios of Gaussians (Shimodaira, 2000) <br>
• Importance-weighting with kernel density estimation (Sugiyama & Müller, 2005)<br>
• Importance-weighting with logistic discrimination (Bickel et al., 2009) <br>
• Kernel Mean Matching (Huang et al., 2006) <br>
• Nearest-neighbour-based weighting (Loog, 2015) <br>
• Transfer Component Analysis (Pan et al, 2009) <br>
• Geodesic Flow Kernel (Gong et al., 2012) (matlab-only)
• Subspace Alignment (Fernando et al., 2013) <br>
• Semi-supervised Subspace Alignment (Yao et al., 2015) (python-only) <br>
• Structural Correspondence Learning (Blitzer et al., 2006) <br>
• Robust Bias-Aware Classification (Liu & Ziebart, 2014) <br>
• Feature-Level Domain Adaptation (Kouw et al., 2016) <br>
• Target Contrastive Pessimistic Risk (Kouw et al., 2017) (python-only)

Python

Installation

Installation can be done through pip:

pip install libtlda

The pip package installs all dependencies. To ensure that these dependencies that don't mess up your current python environment, you should set up a virtual environment. If you're using conda, this can be taken care of by running:

conda env create -f environment.yml
source activate libtlda

Usage

LibTLDA follows a similar structure as scikit-learn. There are several classes of classifiers that can be imported through for instance:

from libtlda.iw import ImportanceWeightedClassifier

With a data set of labeled source samples (X,y) and unlabeled target samples Z, the classifier can be called and trained using:

clf = ImportanceWeightedClassifier(iwe='kmm')
clf.fit(X, y, Z)

Given a trained classifier, predictions can be made as follows:

predictions = clf.predict(Z)

Check the documentation for more information on specific classes, methods and functions.

Matlab

Installation:

First clone the repository and change directory to matlab:

git clone https://github.com/wmkouw/libTLDA
cd libTLDA/matlab/

In the matlab command window, call the installation script. It downloads all dependencies (minFunc, libsvm) and adds them, along with libtlda, to your path:

install.m

Usage

There is an example script that can be edited to test the different classifiers:

example.m

Contact:

Questions, comments and bugs can be submitted in the issues tracker. Any particular method / algorithm / technique that you feel should be included, can be submitted as an issue as well.