This repository contains a set of optimization algorithms and objective functions, and all code needed to reproduce experiments in:
"DESTRESS: Computation-Optimal and Communication-Efficient Decentralized Nonconvex Finite-Sum Optimization" [PDF]. (code is in this file [link])
"Communication-Efficient Distributed Optimization in Networks with Gradient Tracking and Variance Reduction" [PDF]. (code is in the previous version of this repo [link])
Due to the random data generation procedure, results may be slightly different from those appeared in papers, but conclusions remain the same.
If you find this code useful, please cite our papers:
@article{li2022destress,
title={DESTRESS: Computation-Optimal and Communication-Efficient Decentralized Nonconvex Finite-Sum Optimization},
author={Li, Boyue and Li, Zhize and Chi, Yuejie},
journal={SIAM Journal on Mathematics of Data Science},
volume = {4},
number = {3},
pages = {1031-1051},
year={2022}
}
@article{li2020communication,
title={Communication-Efficient Distributed Optimization in Networks with Gradient Tracking and Variance Reduction},
author={Li, Boyue and Cen, Shicong and Chen, Yuxin and Chi, Yuejie},
journal={Journal of Machine Learning Research},
volume={21},
pages={1--51},
year={2020}
}
pip install git+https://github.com/liboyue/Network-Distributed-Algorithm.git
If you have Nvidia GPUs, please also install cupy
.
The gradient implementations of all objective functions are checked numerically.
Linear regression with random generated data. The objective function is
Logistic regression with l-2 or nonconvex regularization with random generated data or the Gisette dataset or datasets from libsvmtools
.
The objective function is
One-hidden-layer fully-connected neural network with softmax loss on the MNIST dataset.