runLlamaModels: Creates a registry which can be used for running several Llama models on a cluster.

Description

It is likely that you need to install some additional R packages for this from CRAN or extra Weka learner. The latter can be done via e.g. WPM("install-package", "XMeans").

Feature costs are added for real prognostic models but not for baseline models.

Usage

runLlamaModels(
  asscenarios,
  feature.steps.list = NULL,
  baselines = NULL,
  learners = list(),
  par.sets = list(),
  rs.iters = 100L,
  n.inner.folds = 2L
)

Value

batchtools registry.

Arguments

asscenarios: [(list of) ASScenario]
Algorithm selection scenarios.
feature.steps.list: [list of character]
Named list of feature steps we want to use. Must be named with scenario ids. Default is to take the default feature steps from the scenario.
baselines: [character]
Vector of characters, defining the baseline models. Default is c("vbs", "singleBest", "singleBestByPar", "singleBestBySuccesses").
learners: [list of Learner]
mlr learners to use for modeling. Default is none.
par.sets: [list of ParamSet]
Param sets for learners to tune via random search. Pass an empty param set, if you want no tuning. Must be in of same length as learners and in the same order. Default is none.
rs.iters: [integer(1)]
Number of iterations for random search hyperparameter tuning. Default is 100.
n.inner.folds: [integer(1)]
Number of cross-validation folds for inner CV in hyperparameter tuning. Default is 2L.