Deep domain adaptation networks (DDAN) is a Python library for domain adapation written in TensorFlow.
The following neural network models are implemented in ddan:
git clone https://github.com/erlendd/ddan.git
cd ddan
sudo pip install .
*OR*
sudo python setup.py install
Full working examples can be found in the examples/ directory.
import ddan
model = ddan.DANNModel(epochs=1000, batch_size=64)
# Xs, ys are for the source domain, Xt, (yt) are for the target domain
model.fit(Xs, ys, Xt)
# Obtain predictions from trained model
yprobs = model.predict_proba(Xval)