msaenet 2.6 (2017-04-23)
- Added two arguments
upper.limits to support coefficient constraints in
msaenet 2.5 (2017-03-24)
- Better code indentation style.
- Update gallery images in
msaenet 2.4 (2017-02-17)
- Improved graphical details for coefficient path plots, following the general graphic style in the ESL (The Elements of Statistical Learning) book.
- More options available in
plot.msaenet() for extra flexibility: it is now possible to set important properties of the label appearance such as position, offset, font size, and axis titles via the new arguments
msaenet 2.3 (2017-02-09)
- Reduced model saturation cases and improved speed at the initialization step for MCP-net and SCAD-net based models when
init = "ridge", by using the ridge estimation implementation from
glmnet. As a benefit, we now have a more aligned baseline for the comparison between elastic-net based models and MCP-net/SCAD-net based models when
init = "ridge".
- Style improvements in code and examples: reduced whitespace with a new formatting scheme.
msaenet 2.2 (2017-02-02)
- Added BIC, EBIC, and AIC in addition to k-fold cross-validation for model selection.
- Added new arguments
tune.nsteps to controls this for selecting the optimal model for each step, and the optimal model among all steps (i.e. the optimal step).
- Added arguments
ebic.gamma.nsteps to control the EBIC tuning parameter, if
ebic is specified by
- Redesigned plot function: now supports two types of plots (coefficient path, screeplot of the optimal step selection criterion), optimal step highlighting, variable labeling, and color palette customization. See
?plot.msaenet for details.
- Renamed previous argument
gamma (scaling factor for adaptive weights) to
scale to avoid possible confusion.
- Reset the default values of candidate concavity parameter
gammas to be 3.7 for SCAD-net and 3 for MCP-net.
- Unified the supported model
family in all model types to be
msaenet 2.1 (2017-01-15)
- Speed improvements in
msaenet.sim.gaussian() by more vectorization when generating correlation matrices.
- Added parameters
epsilon for MCP-net and SCAD-net related functions to have finer control over convergence criterion. By default,
max.iter = 10000 and
epsilon = 1e-4.
msaenet 2.0 (2017-01-05)
- Added support for adaptive MCP-net. See
?amnet for details.
- Added support for adaptive SCAD-net. See
?asnet for details.
- Added support for multi-step adaptive MCP-net (MSAMNet). See
?msamnet for details.
- Added support for multi-step adaptive SCAD-net (MSASNet). See
?msasnet for details.
msaenet.nzv.all() for displaying the indices of non-zero variables in all adaptive estimation steps.
- More flexible
predict.msaenet method allowing users to specify prediction type.
msaenet 1.1 (2016-12-28)
- Added method
coef for extracting model coefficients. See
?coef.msaenet for details.
- New website (https://msaenet.com) generated by pkgdown, with a full set of function documentation and vignettes available.
- Added Windows continuous integration support using AppVeyor.
msaenet 1.0 (2016-09-20)
- Initial version of the msaenet package