FedBN: Federated Learning on Non-IID Features via Local Batch Normalization

This is the PyTorch implemention of our paper FedBN: Federated Learning on Non-IID Features via Local Batch Normalization by Xiaoxiao Li, Meirui Jiang, Xiaofei Zhang, Michael Kamp and Qi Dou

Abstract

The emerging paradigm of federated learning (FL) strives to enable collaborative training of deep models on the network edge without centrally aggregating raw data and hence improving data privacy. In most cases, the assumption of independent and identically distributed samples across local clients does not hold for federated learning setups. Under this setting, neural network training performance may vary significantly according to the data distribution and even hurt training convergence. Most of the previous work has focused on a difference in the distribution of labels. Unlike those settings, we address an important problem of FL, e.g., different scanner/sensors in medical imaging, different scenery distribution in autonomous driving (highway vs. city), where local clients may store examples with different marginal or conditional feature distributions compared to other nodes, which we denote as feature shift non-iid. In this work, we propose an effective method that uses local batch normalization to alleviate the feature shift before averaging models. The resulting scheme, called FedBN, outperforms both classical FedAvg, as well as the state-of-the-art for non-iid data (FedProx) on our extensive experiments. These empirical results are supported by a convergence analysis that shows in a simplified setting that FedBN has a faster convergence rate in expectation than FedAvg.

Usage

Setup

pip

See the requirements.txt for environment configuration.

pip install -r requirements.txt

conda

We recommend using conda to quick setup the environment. Please use the following commands.

conda env create -f environment.yaml
conda activate fedbn

Dataset & Pretrained Modeel

Benchmark(Digits)

• Please download our pre-processed datasets here, put under data/ directory and perform following commands:

  cd ./data
  unzip digit_dataset.zip

• Please download our pretrained model here and put under snapshots/ directory, perform following commands:

  cd ./snapshots
  unzip digit_model.zip

For the original data, please download here or you can directly run the following to download and process data.

cd ./utils
python data_preprocess.py

office-caltech10

• Please download our pre-processed datasets here, put under data/ directory and perform following commands:

  cd ./data
  unzip office_caltech_10_dataset.zip

• Please download our pretrained model here and put under snapshots/ directory, perform following commands:

  cd ./snapshots
  unzip office_caltech_10_model.zip

DomainNet

• Please first download our splition here, put under data/ directory and perform following commands:

  cd ./data
  unzip domainnet_dataset.zip

• then download dataset including: Clipart, Infograph, Painting, Quickdraw, Real, Sketch, put under data/DomainNet directory and unzip them.

  cd ./data/DomainNet
  unzip [filename].zip

• Please download our pretrained model here and put under snapshots/ directory, perform following commands:

  cd ./snapshots
  unzip domainnet_model.zip

Train

Federated Learning

Please using following commands to train a model with federated learning strategy.

• --mode specify federated learning strategy, option: fedavg | fedprox | fedbn

cd federated
# benchmark experiment
python fed_digits.py --mode fedbn

# office-caltech-10 experiment
python fed_office.py --mode fedbn

# DomaiNnet experiment
python fed_domainnet.py --mode fedbn

SingleSet

Please using following commands to train a model using singleset data.

• --data specify the single dataset

cd singleset 
# benchmark experiment, --data option: svhn | usps | synth | mnistm | mnist
python single_digits.py --data svhn

# office-caltech-10 experiment --data option: amazon | caltech | dslr | webcam
python single_office.py --data amazon

# DomaiNnet experiment --data option: clipart | infograph | painting | quickdraw | real | sketch
python single_domainnet.py --data clipart

Test

cd federated
# benchmark experiment
python fed_digits.py --mode fedbn --test

# office-caltech-10 experiment
python fed_office.py --mode fedbn --test

# DomaiNnet experiment
python fed_domainnet.py --mode fedbn --test

Citation

If you find the code and dataset useful, please cite our paper.

@inproceedings{
li2021fedbn,
title={Fed{\{}BN{\}}: Federated Learning on Non-{\{}IID{\}} Features via Local Batch Normalization},
author={Xiaoxiao Li and Meirui Jiang and Xiaofei Zhang and Michael Kamp and Qi Dou},
booktitle={International Conference on Learning Representations},
year={2021},
url={https://openreview.net/pdf?id=6YEQUn0QICG}
}