robscale - Accelerated Estimation of Robust Location and Scale
Estimates robust location and scale parameters using platform-specific Single Instruction, Multiple Data (SIMD) vectorization and Intel Threading Building Blocks (TBB) for parallel processing. Implements a novel variance-weighted ensemble estimator that adaptively combines all available statistics. Methods include logistic M-estimators, the estimators of Rousseeuw and Croux (1993), the Gini mean difference, the scaled Median Absolute Deviation (MAD), the scaled Interquartile Range (IQR), and unbiased standard deviations. Achieves substantial speedups over existing implementations through an 'Rcpp' backend with fused single-buffer algorithms that halve memory traffic for MAD and M-scale estimation, and a unified dispatcher that automatically selects the optimal estimator based on sample size.
Last updated
bootstrapconfidence-intervalsdata-sciencehigh-performance-computingm-estimatormedian-absolute-deviationoutliersparallel-computingrcpprobust-estimationrobust-statisticsscale-estimationsimdstatistical-computingstatisticsonetbbglibccpp
5.31 score 58 scripts 347 downloadscatboostr - Gradient Boosting on Decision Trees
Open-source gradient boosting on decision trees with categorical features support out of the box. A fork of the upstream CatBoost R package packaged for CRAN-style installation; the R API is a strict superset of upstream's catboost.* interface.
Last updated
boostingcatboostcategorical-featurescmakecppdata-sciencedecision-treesensemble-learninggbdtgradient-boostinggradient-boosting-decision-treesmachine-learningpredictive-modelingtree-ensemble
3.18 score 101 scripts