NEWS

---

RLT 6.1.1

• CRAN submission cleanup: fixed vignette pre-building, DESCRIPTION fields, man page examples, and C++ compiler warning.

RLT 6.1.0

• Added OOB (out-of-bag) self-kernel to forest.kernel() via oob = TRUE. Computes kernel co-occurrence only from trees where both observations are OOB, eliminating response-contamination bias for unbiased degrees-of-freedom estimation. Returns Kernel (normalized), N (OOB count), and C (leaf-sharing count).

RLT 6.0.2

• Improved regression distributed variable importance for multiple test samples.
• Fixed small issues in model validation, prediction, and variance-related code paths.

RLT 6.0.1

• Added survival forests with logrank, sup-logrank, and Cox-gradient splitting.
• Added survival linear-combination splitting, observation weights, and survival confidence bands.

RLT 6.0.0

• Rewrote the modeling backend in C++ with Rcpp/RcppArmadillo.
• Added deterministic parallel forests with reproducible random seeds across core counts.

RLT 5.8.0

• Added var.prob for weighted variable sampling across model types.
• Added regression variable-importance variance estimation.

RLT 5.7.0

• Added classification forest support with classification-specific linear-combination splitting.
• Added classification probability prediction and related categorical predictor handling.

RLT 4.2.5

• Added regression linear-combination splitting and reinforcement learning splitting modes.

RLT 4.1.3

• Improved reproducibility, model support, and speed.

RLT 4.0.0

• Migrated the package to Rcpp and RcppArmadillo.

RLT 3.2.3 (2021-09-04)

• Fixed header compatibility issues.

RLT 3.2.2 (2018-08-20)

• Updated classification variable-weight handling.

RLT 3.2.1 (2017-11-19)

• Fixed Solaris compilation issues.

RLT 3.2.0 (2017-11-17)

• Fixed omp.h compilation issues.

RLT 3.1.1

• Fixed a survival split crash.

RLT 3.1.0 (2017-04-09)

• Initial release of the RLT modeling interface.