Skip to contents

Changelog

Source: NEWS.md

maxentcpp 1.0.0

CRAN release: 2026-05-18

  • Initial CRAN release.
  • C++17 implementation of the Maximum Entropy (MaxEnt) species distribution modeling algorithm with R bindings via Rcpp.
  • Core features: linear, quadratic, product, hinge, and threshold feature transformations.
  • Categorical variable support via BinaryFeature class — one binary indicator per distinct level, matching Java Maxent’s BinaryFeature.java. Use the categorical parameter in maxent_generate_features() and maxent_run().
  • Missing data handling: maxent_complete_cases() filters samples with NA, NaN, or NODATA sentinel values (-9999); maxent_impute_means() provides mean-substitution. maxent_run() now automatically removes NODATA samples.
  • Cross-validation via maxent_cross_validate() — stratified k-fold partitioning matching Java Maxent’s SampleSet.splitForCV().
  • Replicate runs via maxent_replicate() — bootstrap or subsample resampling matching Java Maxent’s SampleSet.replicate().
  • Jackknife variable importance via maxent_jackknife() — leave-one-out and only-one variable importance analysis matching Java Maxent’s Runner.jackknifeGain().
  • Spatial projection in raw, logistic, and complementary log-log (cloglog) scales.
  • Terra-native raster I/O with streaming evaluation for low memory usage.
  • Model diagnostics: AUC, variable importance (percent contribution and permutation importance), response curves, and MESS maps.
  • Lambda file I/O compatible with the original Java MaxEnt software.
  • Fixed duplicate C++ and R function definitions that caused compilation errors in certain build configurations.