Domain Adaptive Neural Networks with DJP-MMD

This repository contains codes of the DJP-MMD metric proposed in IJCNN 2020. We later extended its to deep neural networks Domain Adaptive Neural Networks (DaNN) by replacing the marginal MMD in DaNN. Considering this work has not been published, if you are interested in this method, please cite the original paper.

Requirements

• PyTorch (version >= 0.4.1)
• scikit-learn

Experiments

We perform the DJP-MMD in Domain Adaptive Neural Networks in Office-Caltech10 raw images, and this new metric shows better convergence speed and accuracy.

Citation

This code is corresponding to our paper below:

@Inproceedings{wenz20djpmmd,
  title={Discriminative Joint Probability Maximum Mean Discrepancy ({DJP-MMD}) for Domain Adaptation},
  author={Zhang, Wen and Wu, Dongrui},
  booktitle={Proc. Int'l Joint Conf. on Neural Networks},
  year={2020},
  month=jul,
  pages={1--8},
  address={Glasgow, UK}
}

Please cite our paper if you like or use our work for your research, thanks!