/SCSG

Implementation of the SCSG algorithm.

Primary LanguageJupyter Notebook

Implementation of Stochastically Controlled Stochastic Gradient Method

Authors : Marin BOUTHEMY & Kossi NEROMA

This code is an implementation of the Stochastically Controlled Stochastic Gradient Method (SCSG) algorithm designed by Lihua Lei & al..

We have implemented it and test it with a convex and a non-convex problem.

Requirements

The library has some requirements :

  • Python 3
  • Numpy
  • Pandas

To install all this requirement you can run:

pip install -r requirements.txt

Files structure

The library contains the following files: