Skip to content

miki998/D-SGD_varying_topology

Repository files navigation

Description

Our code is largely notebook based to show plots and results relayed in the report. Main points are as follows:

- iid distribution for D-SGD for various topology (part 1)
- non-iid distribution for D-SGD for various topology (part 2)
- non-iid distribution for D-SGD for iteratively modified starting topologies (part 3)
differrentTopology.ipynb     : notebooks computing plots for part 1 and 2
Algo1_findTopology.ipynb     : notebooks computing plots for part 3
Algo1_selectedTopology.ipynb : notebooks computing plots for part 3
experimentsPlot.ipynb        : notebooks computing plots for part 3

Further precised comments can be found within each notebooks

And our report report.pdf summarizes our ideas.

Requirements

Main Library needed

torch
networkx
matplotlib

Getting started

Open jupyter notebooks inside the notebooks folder

jupyter notebook "NOTEBOOK-NAME"

Notebooks can be visited in the order given above in Description.

About

Application of D-SGD on varying topology over time on non-iid datasets (2-label skew)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published