This is the code for paper: FEDIC: Federated Learning on Non-IID and Long-tailed Data via Calibrated Distillation.
Abstract: Dealing with non-IID data is one of the most challenging problems for federated learning. Researchers have proposed a variety of methods to try to eliminate the negative influence of non-IIDness. However, they only focus on the non-IID data but generally assume that the universal class distribution is balanced. In many real-world applications, the universal class distribution is long-tailed such that the model will be highly biased to the head classes and performs poorly on the tail classes, although some methods can alleviate the problem of non-IID data. Therefore, this paper studies the joint problem of non-IID and long-tailed data in federated learning and proposes a corresponding solution called Federated Ensemble Distillation with Imbalance Calibration (FEDIC). To deal with non-IID data, FEDIC uses model ensemble to take advantage of the diversity of models trained on non-IID data. Then, a new distillation method with client-wise and class-wise logit adjustment and calibration gating network is proposed to solve the long-tail problem effectively. We evaluate FEDIC on three long-tailed datasets (CIFAR-10-LT, CIFAR-100-LT, and ImageNet-LT) with the non-IID experimental setting, compared with state-of-the-art methods of federated learning and long-tail learning. Specifically designed experiments also validate the effectiveness of the proposed calibrated distillation method.
The following arguments to the ./options.py file control the important parameters of the experiment.
| Argument | Description |
|---|---|
num_classes |
Number of classes |
num_clients |
Number of all clients. |
num_online_clients |
Number of participating local clients. |
num_rounds |
Number of communication rounds. |
num_data_train |
Number of training data. |
num_epochs_local_training |
Number of local epochs. |
batch_size_local_training |
Batch size of local training. |
server steps |
Number of steps of training calibrated network. |
distillation steps |
Number of distillation steps. |
lr_global_teaching |
Learning rate of server updating. |
lr_local_training |
Learning rate of client updating. |
non_iid_alpha |
Control the degree of non-IIDness. |
imb_factor |
Control the degree of imbalance. |
ld |
Control the trade-off between $L{CE}$ and $\lambda L{KL}.$ |
Here is an example to run FEDIC on CIFAR-10 with imb_factor=0.01:
python main.py --num_classrs=10 \
--num_clients=20 \
--num_online_clients=8 \
--num_rounds=200 \
--num_data_training=49000 \
--num_epochs_local_training=10 \
--batch_size_local_training=128 \
--server_steps=100 \
--distillation_steps=100 \
--lr_global_training=0.001 \
--lr_local_training=0.1 \
--non-iid_alpha=0.1 \
--imb_factor=0.01 \
--ld=0.5