the new function
Err computes error rates directly
the new functions
parameters extract nonzero coefficients, features and parameters respectively.
the package is now more informative. A
print.msgl function has been added and by default a progress bar (using package RcppProgress) is shown when the algorithm is running.
it is now possible to fit models without intercepts by setting
intercept = FALSE.
seed argument in
msgl.cv is deprecated, use
The package now uses the generic sparse group lasso optimizer sglOptim.
A new versioning scheme is used. (package version a.b.c.d is interpreted as follows: a - primary version, b - major updates and fixes, c - source revision as corresponding to R-Forge, d - minor fixes made only to the CRAN branch of the source)