Here we publish our code for the paper Decentralized adaptive clustering of deep nets is beneficial for client collaboration accepted at the International Workshop on Trustworthy Federated Learning in Conjunction with IJCAI 2022 (FL-IJCAI'22).
To run a label shift experiment on CIFAR-10, use the following line.
python main_fed.py --n_rounds 100 --num_clients 100 --local_ep 3 --n_sampled 5 --n_clusters 4 --n_data_train 400 --n_data_val 100 --tau 30 --DAC_var --iid label --dataset cifar10