LoginSign Up
← Back to all sparks
M

mlr3viz

ANALYTICS
Velocity2.5

Visualizations for mlr3 objects including learners, predictions and benchmarks

github.com/mlr-org/mlr3viz ↗

mlr3viz keeps the ecosystem's plots working while the plots themselves move out

mlr3visualizationggplot2roc-curvesr-package
◆Current state
mlr3viz supplies autoplot methods across mlr3 objects — learners, resample and benchmark results, tuning instances, ensemble feature-selection results. Recent releases are mostly defensive: suppressing ggplot2::fortify() warnings on ROC and PRC curves, pinning legend order so plots are deterministic across ggplot2 environments, and tracking mlr3 1.7.2. A visible piece of scope also left, with the LearnerSurvCoxPH plot moving to mlr3proba.
◆Where it's heading
The package is being narrowed toward generic plotting infrastructure while learner-specific plots migrate to the packages that own those learners. What it does add is access rather than new charts — passing parameters through to precrec::autoplot(), better hints when the wrong autoplot type is requested, and a confidence-interval plot for mlr3inferr. Determinism across ggplot2 versions has become a recurring concern, which is what happens when a visualization package is depended on by documentation and tests.
◆Prediction
Following the Cox proportional-hazards precedent, further learner-specific plots are likely to move to their owning packages, leaving mlr3viz with the cross-cutting result objects.

◆Recent moves

  1. 18d ago

    mlr3viz 0.11.1 pins legend order for deterministic plots

    Legend order is pinned in plot_learner_prediction() and in the tuning and optimization autoplot methods so it stays deterministic across ggplot2 environments, alongside mlr3 1.7.2 compatibility. Reproducibility housekeeping.

    View source ↗
  2. 5mo ago

    mlr3viz 0.11.0 quiets ggplot2 fortify warnings on ROC curves

    Suppresses unused-argument warnings from ggplot2::fortify() when drawing ROC and PRC curves under newer ggplot2, and notes that the LearnerSurvCoxPH plot moves to mlr3proba 0.8.4. The migration is the more interesting half, but it lands in the other package.

    View source ↗
  3. 1y ago

    mlr3viz 0.10.1 passes plotting parameters through to precrec

    Parameters can now be passed through to precrec::autoplot() when plotting benchmark and resample results as ROC or PRC curves, a confidence-interval plot arrives for mlr3inferr, and wrong autoplot types produce hints about valid ones. Access to the underlying plotting library rather than new plots of its own.

    View source ↗
  4. 1y ago

    mlr3viz 0.10.0 adds a LearnerSurvCoxPH plot

    Adds a plot for LearnerSurvCoxPH — the same plot that later moves out to mlr3proba, which makes this release a useful marker of where the boundary used to sit.

    View source ↗
  5. 2y ago

    mlr3viz 0.9.0 adds EnsembleFSResult plots

    Adds plots for EnsembleFSResult objects and updates for bbotk 0.9.0 and mlr3tuning 0.21.0. Ensemble feature selection getting a visual treatment is the substantive part.

    View source ↗
  6. 2y ago

    mlr3viz 0.8.0 tracks paradox 1.0.0

    A compatibility release for paradox 1.0.0. Ecosystem tracking with no user-facing change.

    View source ↗