Fits classical sparse regression models with efficient active set algorithms by solving quadratic problems. Also provides a few methods for model selection purpose (cross-validation, stability selection).
0.2-4 (2014-01-16) Minor: + memory leak corrected (sp_mat declaration) + linking to Rcpp/RcppArmadillo headers (requires R 3.0-2)
0.2-3 (2013-08-26) + added back the 'normalize' parameter + standardization is performed within the C++ code + use of sparse conversion from Matrix to Armadillo + corrected bug with the 'intercept' and 'residuals' components of the quadrupen class + added more tests in the inst directory + correction in the documentation + added r.squared to the quadrupen class
0.2-2 (2013-04-08) + minor fix to comply with recent ggplot2 updates.
0.2-1 (2013-02-27) + minor fix to pass CRAN check on Windows operating systems.
0.2-0 (2013-02-26) Major: + added bounded regression (regression penalized by infinity norm + structered l2 norm) + added corresponding functionalies for cross-validation and stability path Minor: + corrected wrong annotations of the stability path (PFER) + handled normalization internally ('normalize' is no longer a parameter) + more simple internal handling of penscales and correction of the rescaling of the intercept + better use of multicore features + handled runtime error exception in RcppArmadillo when the system is singular (end of the solution path) A consequence is quadrupen is less likely to crash due to user's "bad" parametrization + simplification of the C++ code, bugs corrected, probably new ones added :-'( + added 'examples' and 'tests' directories
0.1-0 (2012-10-09) + first build: structured elastic-net with (weighted) quadratic loss, cross-validation and stability selection methods.