NEWS | R Documentation |
Original author and previous maintainer, Martin Vincent, has stopped development. This version is released by a new maintainer. Changes from version 1.3.6 are only to comply to CRAN policies.
Bug fixes
Internal code cleanup, bug fixes and refactoring.
Improved documentation and functionality:
added functions for extracting number of nonzero features and parameters (feature_stat
and parameter_stat
)
Minor updates and documentation corrections
Better handling of R memory protection (fewer PROTECT calls)
Hadling of more complex response types
subsampling uses foreach and doParallel (showing progressbar no longer supported for subsampling and cross validation)
subsampling now supports different lambda sequences for each subsample
Fixes :
fixed sample check bug
minor fixes for some numerical computations
API changes :
all lineaer loss classes must set (a sgl::natural) n_variables, whihch will determine the size of the lineaer preditors ( n_samples x n_variables)
in class MultiResponse n_groups -> n_responses
Improved flexibility and reliability :
added handling of q = 1 cases (each feature has one parameter)
correct handling of non penalized parameters when computing lambda.max (fit non penalized parameters before computing critical lambdas)
increased speed when optimizing with a high number of groups/classes (> 100)
some improvements on error reporting and argument validation
Fixes :
fixed bug related to grouping
fixed seval minor bugs
fixed startup messages