msgl: High Dimensional Multiclass Classification Using Sparse Group Lasso

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.

Downloads: