This is partly the reproduction of the paper of Federated Learning of User Verification ModelsWithout Sharing Embeddings
Only experiments on CelebA is produced by the code.
Note: The scripts will be slow without the implementation of parallel computing.
python>=3.6
pytorch>=0.4
Please only run with python main_uv_alg.py
To change the code length, message length, and d_min safely, open the options file and change the default values.