/federated-learning

Primary LanguagePython

A Unified Linear Speedup Analysis of Stochastic FedAvg and Nesterov Accelerated FedAvg

This repository host the code base for the following paper:

A Unified Linear Speedup Analysis of Stochastic FedAvg and Nesterov Accelerated FedAvg
Zhaonan Qu, Kaixiang Lin, Zhaojian Li, Jiayu Zhou, Zhengyuan Zhou
Arxiv

Run

Example command:

python main.py --dataset=w8a --epsilon=0.1 --num_rounds=500000 --num_epochs=1 --number_user=10 --learning_rate=1 --is_decay=True --lrconst=1554 --machine=gpu --batch_size=4 --num_iteration=50000 --model=binarylogisticregression --optimizer=fedave --dimension=300 --adapt=0 --seed=0 --regularization=1e-05

Reference

If you find this work helpful in your research, please consider citing the following paper. The bibtex are listed below:

Qu, Z., Lin, K., Li, Z., Zhou, J. and Zhou, Z., 2020. A Unified Linear Speedup Analysis of Stochastic FedAvg and Nesterov Accelerated FedAvg. arXiv preprint arXiv:2007.05690.