Added permutation-null filtering for residual-specific IMD; shared filtering continues to tune its cutoff with OOB error.
Made saturation-based entropy selection the default for model and fused top-v, with closed-form fused entropy and the earlier methods retained.
Replaced fork-based parallel paths with PSOCK workers and safer serial defaults for tuning grids.
Reconstructed missing ordinary proximity from tree membership and improved related validation, tests, and documentation.
multiRF 0.2.3
Added the unified plot(fit, type = ...) interface and pkgdown site.
Fixed clustering, OOB weights, IMD, stability, and related edge cases.
Improved forest-fitting speed, parallel safety, and cross-platform reproducibility; seeded unsupervised results change once from 0.2.2.
Added an optional fast cluster_imd() path and simplified internals, dependencies, tests, and documentation.
Relicensed under GPL (>= 3) with RF-SRC provenance and citations.
multiRF 0.2.2
Added plot_cluster_composition() for auditable cluster-by-annotation composition heatmaps with counts, normalized percentages, and consistent package styling.
multiRF 0.2.1
Fix: Corrected the default ytry for supervised multivariate regression.
multiRF 0.2.0
Added multivariate regression and unsupervised forests with parallel fitting.
Partition-based ranger-style tree building for faster splits.
IMD importance computed during tree building (zero-cost).
multiRF 0.1.1
Prepared package metadata for publishing (DESCRIPTION, license, and dependency declarations).
Added a practical README.md with installation and usage instructions.
Added NEWS.md to track user-facing changes.
Hardened build settings to exclude development artifacts.
Removed compiled artifacts from src/ for clean source distribution.