Skip to content

Synchronous implementation and experimentation of a distributed stochastic gradient descent (SGD) used in Support Vector Machines (SVMs) in Spark/Scala

Notifications You must be signed in to change notification settings

adriguerra/hogwild-spark

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

40 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

hogwild-spark

Abstract

The goal of this project is to design, implement and experiment a synchronous version in Spark of a distributed stochastic gradient descent (SGD) used in Support Vector Machines (SVMs) by comparing it with previous synchronous and asynchronous implementations in Python.

The main reference for this project is the Hogwild! paper. The Hogwild! paper is an important paper in the Machine Learning and Parallel Computing community that shows that SGD can be implemented without any locking when the associated optimization problem is sparse. hogwild-python is a synchronous and asynchronous implementation in Python of the Hogwild! algorithm by EPFL students. This project is part of the CS-449 Systems for Data Science course taught at EPFL in the Spring semester of 2019.

About

Synchronous implementation and experimentation of a distributed stochastic gradient descent (SGD) used in Support Vector Machines (SVMs) in Spark/Scala

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published