Skip to contents

Changelog

Source: NEWS.md

lambertReg 0.1.10

  • Add a package logo, a local documentation site, a reproducible example gallery, software citation metadata, and contributor guidance.

  • Streamline the README and organize the introductory vignette as a worked guide to paths, cross-validation, prediction, plots, and diagnostics.

  • Use the previously validated column-contiguous C++ kernel, reusing Lambert roots for penalty and derivative evaluations with unchanged scalar arithmetic, coordinate order and convergence criteria.

  • Remove an unused singular-value decomposition from training preprocessing.

  • Standardize each validation fold once per CV call rather than at every penalty position, using the same training-only statistics.

  • Reuse verified warm-start fits only when the warm and zero initial vectors are exactly identical. Both logical attempt records remain available; attempts$executed and attempts$reused distinguish actual solver calls.

  • Preserve public fitting arguments, the complete coefficient path and CV plot data. Early stopping is not enabled.

lambertReg 0.1.9

  • Added an executable introductory vignette covering preprocessing, cross-validation, prediction, diagnostics, CV plots, coefficient paths and penalty evaluation.
  • Expanded help for returned objects, dimensions, status codes and reproducibility.
  • Added private GitHub installation, citation and support information.
  • Added full vignette builds to the local package-validation workflow.
  • Numerical estimation code, defaults and stored-study results are unchanged.

lambertReg 0.1.8

  • Set CV plot labels, titles and annotation text to black.

lambertReg 0.1.7

  • Introduced blue CV curves, orange selected candidates and compact annotations.

lambertReg 0.1.6

  • Aligned native connected axes with explicit tick limits.

lambertReg 0.1.4

  • Added coefficient-path plots and CV-to-path plot dispatch.

lambertReg 0.1.1

  • Completed the full-training path for CV variable-count displays.

lambertReg 0.1.0

  • Initial fixed-shape Gaussian Lambert regression implementation with regularization paths, fold-specific CV, prediction and numerical diagnostics.