Federated Data Selection with Wasserstein Distance

Stage 1: FL training

Fedprox

python main.py --cnum 0 --n 3 \
               --dataset cifar10 \
               --model simple-cnn \
               --alg fedprox \
               --mu 0.1 \
               --batch_size 64 \
               --lr 0.01 \
               --comm_round 80 \
               --epochs 10 \
               

Description:

• --cnum: cuda num
• --n: number of clients
• --lr: learning rate
• --epochs: local epochs
• --comm_round: global iterations
• --alg: fedavg/fedprox
• --mu: penalty term in fedprox

Stage 2: Projektor

to be updated