NEWS

---

MLwrap 0.3.0 (2025-12-15)

• Depurated bug on classification tasks for Permutation Feature Importance (PFI) estimation

• Added global and pairwise interaction analysis based on Friedman’s H-statistic, including tables and plots of normalized H2 values to inspect overall and pairwise feature interactions across regression and classification models.

MLwrap 0.2.3 (2025-11-11)

• Permutation Feature Importance (PFI) now implemented natively in MLwrap, eliminating the vip package dependency. Output and API remain unchanged.

MLwrap 0.2.2 (2025-11-05)

• Added comprehensive documentation on reproducibility with random seed management guidance across all supported algorithms.

• Released complete tutorial for the MLwrap workflow (arXiv preprint).

• Fixed Integrated Gradients compatibility with torch 0.16.1 through float64 precision implementation.

MLwrap 0.2.1 (2025-10-21)

• Updated to ensure full compatibility and integration with the jamovi framework for module development using MLwrap functionalities.

MLwrap 0.2.0 (2025-10-11)

• Cross validation procedure implemented for Bayesian Optimization (fine tuning)

MLwrap 0.1.1 (2025-09-18)

• Fixed deprecation warnings by replacing aes_string() with aes() and sym() with rlang::sym()
• Corrected plot title assignment in plot_calibration_curve() function
• Minor code improvements for better maintainability

MLwrap 0.1.0 (2025-07-22)

• Initial CRAN release