Tuning spaces from the Kuehn (2018) article.
Source
Kuehn D, Probst P, Thomas J, Bischl B (2018). “Automatic Exploration of Machine Learning Experiments on OpenML.” 1806.10961, https://arxiv.org/abs/1806.10961.
ranger tuning space
num.trees \([1, 2000]\)
replace [TRUE,FALSE]
sample.fraction \([0.1, 1]\)
mtry.ratio \([0, 1]\)
respect.unordered.factors [“ignore”, “order”, “partition”]
min.node.size \([1, 100]\)
splitrule [“gini”, “extratrees”]
num.random.splits \([1, 100]\)
mtry.power
is replaced by mtry.ratio
.
rpart tuning space
cp \([1e-04, 1]\)
maxdepth \([1, 30]\)
minbucket \([1, 100]\)
minsplit \([1, 100]\)
svm tuning space
kernel [“linear”, “polynomial”, “radial”]
cost \([1e-04, 1000]\)
gamma \([1e-04, 1000]\)
tolerance \([1e-04, 2]\)
degree \([2, 5]\)
xgboost tuning space
booster [“gblinear”, “gbtree”, “dart”]
nrounds \([2, 8]\)
eta \([1e-04, 1]\)
gamma \([1e-05, 7]\)
lambda \([1e-04, 1000]\)
alpha \([1e-04, 1000]\)
subsample \([0.1, 1]\)
max_depth \([1, 15]\)
min_child_weight \([1, 100]\)
colsample_bytree \([0.01, 1]\)
colsample_bylevel \([0.01, 1]\)
rate_drop \([0, 1]\)
skip_drop \([0, 1]\)