Personalized Federated Learning for Cross-city Traffic Prediction

Requirements

• Python >= 3.8
• torch = 1.13.0
• numpy = 1.25.2
• tqdm
• sklearn
• scipy

Datasets

Please refer to ST-GFSL

• PEMS-BAY
• METR-LA
• Didi-Chengdu
• Didi-Shenzhen

Training

# run pFedCTP 
python run_pFedCTP.py --algo=pFedCTP --batch_size=32  --target_city=shenzhen --num_rounds=90 --local_epochs=150 --target_epochs=50 --gcn_layers=1
# run pFedCTP-woF
python run_pFedCTP.py --algo=pFedCTP-woF  --batch_size=32  --target_city=shenzhen --num_rounds=90  --local_epochs=150 --gcn_layers=1
# run pFedCTP-Trans
python run_pFedCTP.py --algo=pFedCTP-Trans --batch_size=32  --target_city=shenzhen --num_rounds=90  --target_epochs=50 --gcn_layers=1

Citation

If you find this repository, e.g., the paper, code, and the datasets, useful in your research, please cite the following paper: