Changes in version 0.5.1                        

  - Localized RNG handling in exported simulation and
    stability-selection functions so seed= no longer leaves the caller's
    RNG state modified.
  - Reworked the benchmark generation script so it writes to an explicit
    --output-dir=... path when supplied and otherwise defaults to
    tempdir(), rather than writing into the package tree by default.
  - Updated tests and top-level documentation to match the revised RNG
    and benchmark-output behavior.

                 Changes in version 0.5.0 (2026-04-10)                  

  - Added pkgdown website and a package-style README.
  - Added targeted FDA benchmark sensitivity utilities with
    run_selectboost_sensitivity_study().
  - Added simulation controls for confounding_strength,
    active_region_scale, and local_correlation so benchmarks can stress
    the settings where FDA-aware grouping is expected to help.
  - Added shipped benchmark artifacts under inst/extdata/benchmarks/,
    including feature-level mean F1 summaries and ranked
    selectboost_fda() versus plain SelectBoost settings.
  - Added a reproducible benchmark script in
    tools/run_selectboost_sensitivity_study.R and updated the benchmark
    vignette to read the saved study outputs directly.

                        Changes in version 0.4.0                        

  - Added minimal examples to the core functions of the package
  - Added a validation layer with plain_selectboost(),
    simulate_fda_scenario(), evaluate_selection(),
    benchmark_selection_methods(), and run_simulation_study().
  - Added mapped ground-truth utilities so feature-, group-, and
    basis-level recovery can be evaluated on transformed FDA designs.
  - Added a simulation and benchmarks vignette plus release-hardening
    metadata for CI and pkgdown workflows.

                        Changes in version 0.3.0                        

  - Added a broader selector interface with lasso, group_lasso, and
    sparse_group_lasso aliases, while keeping backend-specific names
    available.
  - Added sparse-group lasso support through the SGL package.
  - Added overlapping interval groups and region-aware association
    structures for FDA grouping.
  - Added calibration helpers for stability-selection parameters,
    interval widths, and SelectBoost c0 grids.
  - Added method-comparison utilities to run grouped stability
    selection, interval stability selection, FDA-SelectBoost, and
    optional FDboost workflows on the same fda_design.
  - Added a formula interface with fda_design_formula(),
    fit_stability_formula(), and fit_selectboost_formula().

                        Changes in version 0.2.0                        

  - Added FDA-native preprocessing objects for identity transforms,
    scalar standardization, spline-basis expansion, and FPCA.
  - Added fitted preprocessing workflows with fit_fda_preprocessor() and
    apply_fda_preprocessor() so training and new-data transforms use the
    same mapping.
  - Extended fda_design() to support multiple functional predictors,
    scalar covariates, optional fitted preprocessors, and richer
    reversible domain metadata.
  - Standardized fit outputs across stability selection and SelectBoost
    with consistent print(), summary(), selection_map(), plot(), and
    selected() behavior.
  - Added packaged example datasets for end-to-end workflows and updated
    the vignettes to start from raw functional inputs.
  - Expanded test coverage and refreshed package documentation for the
    FDA-native core API.

                        Changes in version 0.1.0                        

  - Initial package release.
  - Added grouped stability selection for functional predictors
    represented on grids or in basis form.
  - Added FDA-aware SelectBoost wrappers, interval grouping helpers,
    plotting methods, and introductory vignettes.