glossary

library(daisugi)

daisugi is an experimental tree-machine curation repository focused on unconventional forests, probabilistic boosters, interpretable ensembles, and emerging recursive partition systems.

This glossary acts as:

a machine registry
a terminology guide
a reference index
a research companion

Most engines in daisugi expose a lightweight “(x, y)” interface and follow a consistent verb-oriented workflow.

Core Verbs

daisugi standardizes machine workflows through a small set of forestry-inspired verbs.

“grow_*()”: fit or train a machine
“harvest_*()”: generate predictions
“trees": unified argument for rounds / iterations / estimators

Machine Registry

The following registry links implemented engines to their originating projects, papers, or repositories. Registy also lists the alias in daisugi as well as a set of haikus for TL/DR referencing.

	link	alias	haiku
boulevard	https://github.com/siriuz42/boulevard	boulevard	boosted trees converge where the kernel always knew the road was a proof
cforest	https://partykit.r-forge.r-project.org/partykit/	conditional	not gain, but a test the split earns its place through chance bias cannot root
ebm	https://interpret.ml/docs/ebm.html	explainable	each feature speaks alone the sum is the whole model darkness becomes light
evtree	https://cran.r-project.org/web/packages/evtree/index.html	evolutionary	natural selection the whole tree evolves at once no greedy shortcuts
nrgboost	https://github.com/ajoo/nrgboost	energy	non-discriminative sample all columns with ease no target required
ngboost	https://github.com/stanfordmlgroup/ngboost	natural	not a point, but cloud the natural curve of space to show all outcomes
perpetual	https://perpetual-ml.github.io/perpetual/	perpetual	one knob, the budget self-generalizing design no search, just descent
snapboost	https://ibmsoe.github.io/snap-ml-doc/v1.6.0/manual.html#snapboost	snap	depth rolls like dice decision or linear random number tricks
wildwood	https://github.com/pyensemble/wildwood	wild	every pruning weighed out-of-bag shadows voting wildest of all
yggdrasil	https://github.com/google/yggdrasil-decision-forests	yggdrasil	sacred trunk rises c++ decision trees fast quiet giant

Supported Tasks

Machine	Classification	Regression
Boulevard	❌	✅
Conditional Trees	✅	✅
EBM	✅	✅
Evolutionary Trees	✅	✅
NGBoost	✅	✅
NRGBoost	✅	✅
Perpetual	✅	✅
SnapBoost	✅	✅
WildWood	✅	✅
Yggdrasil	✅	✅

tidymodels interface

Machine	Supported?
EBM	✅
Yggdrasil	✅

EBM is supported as a boost_tree as well as an explainable_boost model engine. This allows lite-weight as well as additional control over the model specification & hyperparameter optimization. daisugi contains the additional dials necessary for tuning EBMs. Read more about this in the tidymodels vignette.