NEWS.md

In funcml: Functional Machine Learning Framework

funcml 0.9.0

• Prepared for CRAN resubmission: removed the unused methods package from Imports, added explicit importFrom declarations for predict(), model.frame(), model.response(), setNames(), median() (stats) and tail() (utils) via a new package-level R/funcml-package.R, and excluded LICENSE.md from the build (the package uses the standard GPL-3 license text that R bundles automatically; the file remains in the GitHub repo only). R CMD check --as-cran now passes with 0 errors, 0 warnings, 0 notes.

funcml 0.8.9

• Renamed compare_learners() to compare() for consistency with the other short, verb-named entry points (fit(), evaluate(), tune(), interpret(), estimate()). This is a breaking change with no backward-compatible alias: code calling compare_learners() must be updated to call compare(). Note that compare() masks testthat::compare() when both packages are attached.

funcml 0.8.8

• Prepared for CRAN resubmission (previous CRAN version: 0.7.1): removed the redundant Author/Maintainer DESCRIPTION fields (they had drifted out of sync with Authors@R, triggering a R CMD check NOTE) in favor of deriving them from Authors@R, and fixed a non-canonical CRAN task view URL in the README. R CMD check --as-cran now passes with 0 errors, 0 warnings, 0 notes.

funcml 0.8.7

• Removed the stale hand-written inst/CITATION, which hardcoded "R package version 0.7.1" and had drifted six releases behind. R now falls back to the default citation auto-generated from DESCRIPTION, which always reflects the installed version.
• README's Citation section now shows the live citation("funcml") output instead of hardcoded text.

funcml 0.8.6

• Reverted the SHAP/local-surrogate bar plot colors to green = positive, red = negative (the original convention), after the two intermediate swaps in 0.8.1 and 0.8.5.

funcml 0.8.5

• Added numeric value labels to every bar-chart-style plot for easier reading: SHAP waterfall (+0.148/-0.089 style, signed), SHAP feature importance, SHAP interaction strength, and local surrogate (interpret(method = "local_model")) contributions. The local surrogate plot's colors were also swapped to match the SHAP convention (positive = red, negative = green).

funcml 0.8.4

• Fixed a hang introduced in 0.8.2: interpret(method = "shap", ncores = <n>) with xgboost, lightgbm, mlp, densemlp, or bart on Unix could hang indefinitely, because functionals::fmap() forks the process (parallel::mclapply()) and those models' fitted state holds a C/C++ handle that is not valid in the forked child. ncores is now ignored (with a warning) for those models, falling back to sequential; other models parallelize as before.

funcml 0.8.3

• SHAP waterfall (kind = "waterfall") is now a zero-anchored per-feature contribution bar chart instead of a cumulative chained waterfall: every bar starts at 0 and extends to its own SHAP value, so no bar crosses from one side of the reference line to the other. The vertical reference line is fixed at 0 instead of the baseline prediction.
• SHAP beeswarm/summary (kind = "summary"/"beeswarm") now uses the standard SHAP blue (low) to red (high) colorbar on the right, instead of the previous bottom yellow-to-purple legend, and drops the per-feature numeric labels for plain feature names. It also gained a v argument to restrict the plot to a single feature.

funcml 0.8.2

• interpret(method = "shap") gained an ncores argument that parallelizes the per-observation Monte Carlo SHAP computation via functionals::fmap() (the same backend already used by evaluate()/tune()/compare_learners()). Each observation is now seeded independently (seed + observation_index - 1) so results are identical whether run sequentially or in parallel; this changes the exact values produced by a seeded interpret(method = "shap") call compared to earlier releases, though the estimator itself (Monte Carlo permutation SHAP) is unchanged.

funcml 0.8.1

• Printed result tables (evaluate(), compare_learners(), tune(), interpret(method = "calibration"), interpret(method = "dca"), roc_curve(), auc_ci()) now round numeric columns to 4 digits by default (digits argument on the relevant print()/summary() methods and on auc_ci()), instead of printing full floating-point precision.
• Swapped the SHAP waterfall/force colors: positive contributions are now red, negative are green.
• Reworked the SHAP beeswarm/summary plot (kind = "beeswarm" / "summary") to show mean |SHAP value| next to each feature name, a yellow-to-purple viridis "plasma" feature-value gradient, and a bottom legend with Low/High endpoints. Also fixed a row-order misalignment bug in the per-feature value scaling introduced by the 0.8.0 native SHAP plot rewrite.

funcml 0.8.0

• Added densemlp as a new learner, wrapping the published densemlp CRAN package. It complements the existing built-in mlp learner with richer architecture options (residual connections, gated blocks, input projection, focal loss, label smoothing, LR schedules) for regression and classification.
• Added roc_curve() and auc_ci(), backed by the pROC package: roc_curve() returns the full sensitivity/specificity curve plus a plot() method, and auc_ci() reports AUC with a DeLong (default) or bootstrap confidence interval. funcml's own fast auc() is unchanged and remains what resampling/tuning use internally.
• Added decision curve analysis (Vickers and Elkin, 2006): dca() computes net benefit across risk thresholds for the model, "treat all", and "treat none" strategies, and interpret(method = "dca") runs it directly on a fitted binary classifier with a plot() method.
• Removed the shapviz dependency. All SHAP plot kinds (waterfall, force, summary/beeswarm, importance/bar, dependence, dependence2d, interaction) are now native ggplot2 implementations reading directly from funcml's own SHAP result table. The underlying SHAP values were already funcml's own Monte Carlo permutation estimate (interpret(method = "shap")); shapviz was only ever used for plotting.
• Reworked theme_funcml() to match the CLAVUS Nature Medicine figure style: theme_classic() base, Okabe-Ito colorblind-safe palette, bold unboxed strip labels, and grey92 major gridlines. All package plots (interpret(), evaluate(), compare_learners(), tune(), estimate()) now share this theme instead of each building its own ad-hoc theme_bw()/theme_minimal() variant.
• Migrated internal row-accumulation (resampling folds, tuning grids, learner comparisons, PDP/ICE/ALE curves, permutation importance, interaction grids, MLP training history) from do.call(rbind, ...) to data.table::rbindlist() for faster combination of many small result frames. All public return objects remain plain data.frames; no API or behavior change.
• Added a citation for Naimi, Cole, and Kennedy (2016) to DESCRIPTION, covering the plug-in g-computation method.
• plot.funcml_pdp() now fixes the y-axis to the [0, 1] probability scale for classification PDPs (type = "prob"), instead of auto-scaling to the local range of the curve, which could visually exaggerate small effects. Regression PDPs are unaffected.
• Moved every learner engine package (MASS, mgcv, nnet, rpart, glmnet, ranger, e1071, randomForest, gbm, C50, kknn, earth, naivebayes, mda, ada, pls, partykit, dbarts, torch, xgboost, lightgbm, densemlp) from Suggests to Imports, so a standard installation always has every advertised learner available and learners()/fit() cannot fail with a missing-package error for a registered model.

funcml 0.7.2

• Added mlp as an internal torch-backed learner for regression, binary classification, and multiclass classification.
• Added CRAN installation instructions to the README and kept the GitHub installation path for development snapshots.
• Added a README note that the funcml companion paper is submitted to JMLR.

funcml 0.7.1

• Refined the README into a more detailed progressive API walkthrough with additional tables, figures, and staged examples covering the full package surface.
• Hardened interpretability runtime paths by forcing vip to use permutation importance consistently while retaining shapviz-enhanced SHAP plotting when the optional plotting packages are installed.

funcml 0.7.0

• Consolidated funcml as a machine learning framework for R with stable S3 interfaces for fitting, prediction, evaluation, tuning, learner comparison, interpretation, and plug-in g-computation.
• Added richer resampling support through plain holdout, grouped cross-validation, and time-aware rolling splits.
• Added uncertainty summaries to evaluate() and compare_learners(), including fold-level standard errors and confidence intervals in summaries and plots.
• Added random-search tuning with search = "random" and n_evals, plus nested resampling support in tune() for outer-fold performance estimates of the model-selection procedure.
• Hardened the fit/predict contract with clearer errors for missing predictor columns and unseen factor levels, stricter probability-output normalization, and broader learner contract coverage across the registry.
• Added multiclass and weighted AUC support and clarified default evaluation behavior for binary versus multiclass classification.
• Added list_learners() as a learner capability catalog and improved package metadata, citation, and repository scaffolding for release and paper preparation.
• Removed the catboost learner backend from the registry and package metadata.
• Kept lightgbm as a standard learner dependency available with funcml.

funcml 0.2.0

• Added richer evaluation-centered resampling with plain holdout, grouped cross-validation, and time-aware rolling splits.
• Added uncertainty summaries to evaluate() and compare_learners(), including fold-level standard errors and confidence intervals in summaries and plots.
• Extended estimate() with configurable interval reporting, including bootstrap percentile intervals for average causal estimands.
• Added random-search tuning with search = "random" and n_evals for budgeted hyperparameter search.
• Added nested resampling to tune() via outer_resampling, so tuning can report unbiased outer-fold performance estimates for the selected workflow.
• Hardened the fit/predict contract with clearer errors for missing predictor columns and unseen factor levels, plus stricter probability-output normalization.
• Expanded the test suite with focused coverage for resampling, uncertainty, tuning, and prediction-contract behavior.

funcml 0.1.1

• Vendored canonical interpretability implementations from vip, pdp, iml, and a minimal internal shapviz layer.
• Replaced runtime vip and pdp dependencies with internal implementations while preserving the existing funcml entrypoints.
• Added parity tests against sourced upstream reference code for permutation importance, PDP, ICE, ALE, Shapley values, and local surrogate explanations.
• Switched local / local_model to an iml::LocalModel-style sparse local surrogate using glmnet and Gower weighting.

funcml documentation built on Aug. 22, 2026, 5:08 p.m.