Crate multistochgrad[−][src]

Modules

applis

Some applications providing examples and tests. They show how implement Traits defined in types.rs to build a use case.

mnist

Structure and functions to read MNIST database

prelude

to include necessary files

sag

Implementation of basic Stochastic Average Gradient (SAG)

scsg

A Rust implementation of Lei-Jordan papers:
"On the adaptativity of Stochastic gradient based optimization" (2019) SCSG-1
"Less than a single pass : stochastically controlled stochastic gradient" (2019) SCSG-2

svrg

A Rust implementation of Johnson-Zhang paper:
“Acceleration stochastic Gradient Descent using Predictive Variance Reduction”
SVRG

types

This file is inspired by the crate optimisation written by Oliver Mader b52@reaktor42.de.
I kept the traits Function, FunctionC1, Summation and SummationC1 which provides the interface for users to define a minimisation problem.