CWAN

Multi-source Heterogeneous Domain Adaptation with Conditional Weighting Adversarial Network

Running Environment

Prerequisites

• python 3.6
• tensorflow-gpu 1.4
• CUDA 8.0
• cudnn 6.0
• numpy
• scipy
• matplotlib
• scikit_learn

Step-by-step Installation

conda create -n cwan python=3.6
conda activate cwan

pip install tensorflow-gpu==1.4
conda install cudatoolkit=8.0
conda install cudnn=6.0
conda install scipy
conda install matplotlib
conda install scikit-learn

Datasets

You can download the example dataset from here (Password: 8i7c), and put in the folder of datasets.

All of the datasets can be downloaded from here (Password: 4y1q).

Running

1. You can run this code by inputing:

python -W ignore main.py

The results should be close to 59.67 (A (D_{4096}), D (R_{2048}) -> W (S_{800})). Note that different environmental outputs may be different.

2. You can use your datasets by replacing:

source_exp = [ad.SAD, ad.SDR]
target_exp = [ad.TWS]

3. You can tune the parameters, i.e., lr_1 (learning rate for g(\dot), f(\dot)), lr_2 (learning rate for d(\dot)), T, d, beta, tau, for different applications.

4. The default parameters are: lr_1 = 0.004, lr_2 = 0.001, T = 500, d = 256, beta = 0.03, tau = 0.004.

Citation

If you find this helpful, please cite:

@ARTICLE{9530273,
  author={Yao, Yuan and Li, Xutao and Zhang, Yu and Ye, Yunming},
  journal={IEEE Transactions on Neural Networks and Learning Systems}, 
  title={Multisource Heterogeneous Domain Adaptation With Conditional Weighting Adversarial Network}, 
  year={2021},
  volume={},
  number={},
  pages={1-14},
  doi={10.1109/TNNLS.2021.3105868}}