NEWS.md

In agisga/grpSLOPE: Group Sorted L1 Penalized Estimation

grpSLOPE (development version)

This is a list of software changes.

grpSLOPE 0.3.3

• No API changes
• Very minor changes based on feedback from CRAN review.

grpSLOPE 0.3.2

• No API changes
• Only changes are to the automated software tests to accommodate minor floating point arithmetic differences on M1mac

grpSLOPE 0.3.1

• Only changes are to the automated software tests.

grpSLOPE 0.3.0

• Removes the dependency on the R package SLOPE, because the only two functions from SLOPE that are used in grpSLOPE are being deprecated and removed from SLOPE package versions newer than 0.1.3.
• Since the dependency on the R package SLOPE has been removed (see above), the two functions SLOPE::SLOPE_solver and SLOPE::prox_sorted_L1 (including the underlying C implementation) have been adapted from SLOPE version 0.1.3 into this version of the grpSLOPE package.
• In addition to the default FISTA solver (function proximalGradientSolverGroupSLOPE used by default within function grpSLOPE), an ADMM solver has been implemented for the Group SLOPE model (function admmSolverGroupSLOPE).

grpSLOPE 0.2.1

• Fix for the installation error on r-oldrel-windows-ix86+x86_64 (R v3.2.5) in the CRAN package check results. This error was caused by the generic S3 method sigma() not being available from the stats package prior to R v3.3.0.

• Checks for missing data in the inputs X, y and group were added in grpSLOPE().

• A check for whether the input matrix X has columns with 0 variance is performed in grpSLOPE() when normalize=TRUE. This should prevent division by 0, when the columns of X are standardized to have norms equal to 1.

agisga/grpSLOPE documentation built on June 3, 2023, 10:16 a.m.