This is the code for paper "Safe Deep Semi-Supervised Learning for Unseen-Class Unlabeled Data" published in ICML 2020.
The code is implemented with Python and Pytorch.
Here is an example:
python train.py --dataset MNIST --ratio 0.6 --n_labels 60 --iterations 200000
We thank the Pytorch implementation on Meta-Net (https://github.com/xjtushujun/meta-weight-ne) and learning-to-reweight-examples(https://github.com/danieltan07/learning-to-reweight-examples).
If you have any questions, please contact Lan-Zhe Guo ([email protected]).