NEWS
BKP 0.3.0 (2026-07-02)
- Added
fit_TwinDKP() and associated S3 methods for scalable global-local Dirichlet Kernel Process modeling.
- Added
fit_TwinBKP() with full S3 support for scalable Twin Beta Kernel Process modeling.
- Implemented Twinning-based global subset selection and kd-tree local-neighbour search.
BKP 0.2.4
- Improved computational efficiency by implementing kernel evaluation, prior construction, loss evaluation, and hyperparameter optimization routines in C++.
- Added support for the compactly supported Wendland kernel via
kernel = "wendland".
- Added optional Shepard effective-sample-size calibration for
fit_BKP(ess = "shepard") and fit_DKP(ess = "shepard"), while keeping the default ess = "none" behavior unchanged.
- Added ggplot2 support for plotting;
plot.BKP(..., engine = "ggplot") now produces ggplot2-based visualizations.
- Added a package-wide
isotropic argument, defaulting to TRUE, for isotropic kernels with a shared length-scale across dimensions. Set isotropic = FALSE to use anisotropic kernels with dimension-specific length-scales.
BKP 0.2.3 (2025-09-22)
- Removed vignettes to avoid redundancy with the arXiv paper and to resolve CRAN thread limit issues during vignette building.
BKP 0.2.0 (2025-09-16)
- Added
fitted(), parameter(), and quantile() methods.
- Updated
predict() and simulate() methods: both now return results for the training data by default when Xnew is not provided.
- Extended
plot() method with new dims argument for higher-dimensional inputs.
- Added Section 5 to the vignette, presenting a real-data application on Loa loa parasite infection in North Cameroon.
- Added argument checking with informative error messages.
BKP 0.1.1 (2025-08-19)
- Added a vignette introducing the package.
- Fixed minor bugs and improved stability.
BKP 0.1.0 (2025-07-23)