Learn R Programming

⚠️There's a newer version (2.3.9) of this package.Take me there.

msgl (version 0.1.3)

High dimensional multiclass classification using sparse group lasso

Description

Sparse group lasso multiclass classification, suitable for high dimensional problems with many classes. Fast algorithm for solving the multinomial sparse group lasso convex optimization problem. This package apply template metaprogramming techniques, therefore -- when compiling the package from source -- a high level of optimization is needed to gain full speed (e.g. for the GCC compiler use -O3). Use of multiple processors for cross validation and subsampling is supported through OpenMP. The Armadillo C++ library is used as the primary linear algebra engine. Armadillo is licensed under the MPL 2.0. The Armadillo C++ library is primarily developed at NICTA (Australia) by Conrad Sanderson, with contributions from around the world. Furthermore the package utilize various Boost libraries, in particular the Tuple library by Jaakko Jarvi and the Random library by Jens Maurer. The Boost libraries are licensed under the Boost Software License.

Copy Link

Version

Install

install.packages('msgl')

Monthly Downloads

154

Version

0.1.3

License

GPL (>= 2)

Maintainer

Martin Vincent

Last Published

May 27th, 2013

Functions in msgl (0.1.3)

predict.msgl

Predict
msgl.cv

Multinomial sparse group lasso cross validation using multiple possessors
msgl.lambda.seq

Computes a lambda sequence for the regularization path
sim.data

Simulated data set
msgl

Fit a multinomial sparse group lasso regularization path.
sgl.standard.config

Standard algorithm configuration
sgl.algorithm.config

Create a new algorithm configuration
msgl.subsampling

Multinomial sparse group lasso generic subsampling procedure