Added new algorithm: fitted group lasso.

Added new algorithm: fitted sparse-group lasso.

Added new parameter: l2_fitted_values to enable the new algorithms.

Added new parameter: step_size - mandatory for the fitted sparse-group lasso.

Added new parameter: delta - mandatory for both new algorithms.

Added new parameter: standardize to enable automated standardization of input data.

Imported the R package “matrixStats” since it is used for the standardization.

Updated the help pages and vignette to provide information about the new algorithms and parameters.

Deleted “trace_progress = T” in all examples, since it caused –run-donttest to fail.

Added a reference in the package documentation.

- Wrapped the entire example section in seagull.R into a environment to circumvent troubles with the Solaris OS.

Replaced the pre-calculation of the matrix

`X^T X`

, by matrix-vector multiplications for the gradients in src_lasso.cpp, src_group_lasso.cpp, and src_sparse_group_lasso.cpp. This circumvents potential memory issues for the allocation of`X^T X`

.Fixed an issue related to the variable TEMP2 in src_lasso.cpp, src_group_lasso.cpp, and src_sparse_group_lasso.cpp

Replaced

`\dontrun`

by`\donttest`

in R.R.Shortened title in DESCRIPTION.

- Fixed a limitation for the design matrix X. The matrix may now have
more columns than rows. But each default algorithm to calculate
`max_lambda`

will fail, because the inverse of`X^T X`

is explicitly needed. However, if a value for`max_lambda`

is provided upon calling the function`seagull`

, a solution will be calculated.

Added parameter

`trace_progress`

. Default is`FALSE`

.Added general vignette.

- Exchanged wrappers
`seagull_lasso`

,`seagull_group_lasso`

, and`seagull_sparse_group_lasso`

by`seagull`

. The different penalties shall now be called by specifying the mixing parameter`alpha`

. This parameter was initially only necessary for the sparse-group lasso. But the lasso and the group lasso are limiting cases, where`alpha = 1`

and`alpha = 0`

, respectively. So, now both regularizations may be initialized by calling the function`seagull`

with`alpha = 1`

or`alpha = 0`

.

- Added documentation.

- Added different exemplary data set (
`seagull_data`

).

Added wrapper

`seagull_lasso`

for the`seagull_lasso_Rcpp.cpp`

Added wrapper

`seagull_group_lasso`

for the`seagull_group_lasso_Rcpp.cpp`

Added wrapper

`seagull_sparse_group_lasso`

for the`seagull_sparse_group_lasso_Rcpp.cpp`