This is a demo for the paper: "RECORD: Resource Constrained Semi-Supervised Learning under Distribution Shift".
- Python 3.7
- Pytorch
- scikit-learn
- h5py
- matplotlib
- src: Python source code.
- data: Placeholder for the dataset. Please download the dataset from Google Drive.
- logs: Placeholder for the running logs.
- images: Placeholder for the line charts.
For example, you can run the following command in the root path:
python ./src/run.py
The result will be saved in logs folder with a line chart saved in the images folder.
In this demo, we prepared four benchmark data sets for the distribution shift and three implements of semi-supervised learning methods.
You can also use command python ./src/run.py -h
to list the usages.
Contact wnjxyk@gmail.com for more questions.