sgdnet

Sparse Linear Models for Big Data via Stochastic Average Gradient

Johan Larsson https://larssonjohan.com/ (Department of Statistics, Lund University) , Toby Dylan Hocking https://tdhock.github.io/ (Northern Arizona University) , Michael Weylandt https://michaelweylandt.github.io (Department of Statistics, Rice University)

The sgdnet R package implements a stochastic gradient method for elastic-net penalized generalized linear models. The package is stable and correct, but performance is not quite what we hoped to attain. Researchers seeking to implement stochastic gradient methods for regularized linear models are advised to take advantage of modern deep learning frameworks and massively parallel computational hardware instead.

Direct Link: https://github.com/jolars/sgdnet

Package Documentation: http://jolars.github.io/sgdnet/