Home

Awesome

JDOT

Joint distribution optimal transportation for domain adaptation

imgreg

This repository contains the code used for the numerical experiments of the foillowing paper:

N. Courty, R. Flamary, A. Habrard, A. Rakotomamonjy, "Joint Distribution Optimal Transportation for Domain Adaptation", Neural Information Processing Systems (NIPS), 2017.

The code is under MIT Licence but please refer to and cite the above paper if you use it for academic purposes.

Note that this code has not been thouroughly tested and a more clean/robust implementation will be added to the POT Toolbox that already have numerous optimal transport domain adaptation methods.

Dependencies

In order to run, the code requires the following Python modules:

If you want to use the neural network JDOT example you will also need the keras toolbox.

If you have not already installed them you can install the dependencies with PIP using the following command

$ pip install numpy scipy matplotlib POT # keras

Modules

Plot examples

We provide 3 example scripts that reproduce the Figures in the paper.

imgreg

imgreg