Changelog

Source: NEWS.md

daisugi 0.0.9

• add the three well-known boosting algorithms with a twist
  • catboost based on the langevin boosting methodology
  • lightgbm with the GBDT-PL methodology
  • xgboost as a boosted forest
  • add gaiden.Rmd

daisugi 0.0.8

• clean up daisugi.Rmd

• improve tidymodels implementations

  • add parsnip S3 methods for explainable_boost
  • modify prediction bridges for ebm & ydf
    • output tibbles for class & numeric

• add parent class for all machines (preparing for future methods)

  • daisugi_mother class applied to all fit objects

daisugi 0.0.7

• initial tidymodels bindings for ebm & ydf
  • parsnip register & added special dials hyperparameters
    • ebm contains bindings to boost_tree and a new custom explainable_boost which contains new dials.
    • refactor fitting and predicting within yggdrasil to handle matrices passed from parsnip
• improving documentation
  • small typos & fixes
  • splitting up documentation into parts:
    • daisugi.Rmd - Getting Started with Classification
    • regression.Rmd - Extended Documentation for Regression
    • probabilistic.Rmd - TBD, Prediction Intervals with-and-without daisugi
    • tidymodels.Rmd - ‘ebm’ & ‘ydf’ engine bindings.
    • super.Rmd - TBD, a super learner, built with daisugi, for daisugi

daisugi 0.0.6

• add support for Boulevard
  • fit with grow_boulevard_trees
  • predict with harvest_boulevard_trees
  • “Regularized Stochastic Gradient Boosted Trees and Their Limiting Distribution”

daisugi 0.0.5

• improve documentation

• add support for cforest

  • fit with grow_conditional_trees
  • predict with harvest_conditional_trees
  • a ctree: “statistical approach [to recursive partitioning] which takes into account the distributional properties of the measures”
    • cforest is a bagging ensemble of ctrees

• add support for evtree

  • fit with grow_evolutionary_trees
  • predict with harvest_evolutionary_trees
  • genetic tree optimization technique

• add support for NRGBoost

  • fit with grow_energy_trees
  • predict with harvest_energy_trees
  • a tree-based generative algorithm for tabular data
    • “Unlike discriminative methods, NRGBoost can be used to predict any column in the data, not just a specific”target” column.”

daisugi 0.0.4

• added support for NGBoost
  • fit with grow_natural_trees
  • predict with harvest_natural_trees
  • a natural gradient booster from Stanford ML Group
• added documentation for package.
  • daisugi.Rmd
    • getting started vignette
  • glossary.Rmd
    • a note on machines & nomenclature
• added images for vignettes and README documentation

daisugi 0.0.3

• added support for EBMs
  • fit with grow_explainable_trees
  • predict with harvest_explainable_trees
  • a ‘glassbox’ novel gam boosting machine

daisugi 0.0.2

• added support for WildWood
  • fit with grow_wild_trees
  • predict with harvest_wild_trees
  • a newer and advanced random forest algorithm
• added support for Perpetual
  • fit with grow_perpetual_trees
  • predict with harvest_perpetual_trees
  • a budget-based autoML-type tree machine

daisugi 0.0.1

• added support for Yggdrasil Decision Forest
  • fit with grow_yggdrasil_trees
  • predict with harvest_yggdrasil_trees
  • YDF Gradient Boosted Trees (Google)
• added support for SnapBoost
  • fit with grow_snap_trees
  • predict with harvest_snap_trees
  • Snap ML’s Boosting Machine (IBM)