arcfsAggregator: Aggregator for the arc-fs algorithm.

In boostr: A modular framework to bag or boost any estimation procedure.

Description

A (parallelized) implementation of the aggregator described in the arc-fs algorithm.

Usage

arcfsAggregator(estimators, beta, ..., .parallelPredict = FALSE,
  .parallelTally = FALSE, .rngSeed = 1234)

Arguments

estimators	a list of estimators which must produce output in the same response-space. This is usually the output of some reweighter function.
...	this does nothing – meant to swallow auxillary output from reweighter function.
.parallelPredict	a boolean indicating if prediction should be carried out in parallel.
beta	a vector of scalar weights associated to each estimator in estimators
.parallelTally	a boolean indicating if vote tallying should be performed in parallel. Unless you have more than 1,000 votes / observation, you probably won't see much performance gain by parallelizing this step.
.rngSeed	the RNG seed sent to predictClassFromWeightedVote in the case of a tie.

Details

By default, this function will perform its predictions in sequence across the estimators in estimators. To predict in parallel, change .parallelPredict to TRUE.

Value

a function whose sole argument is newdata and whose output is the aggregated predictions of the boosted ensemble, estimators.

For internal bookkeeping, this function is inherits from the 'aggregator' class.

Note

In accord with the arc-fs algorithm, there is the assumption that the estimators in estimators are classifiers. More aptly, their output is either of factor or character-type.

See Also

Other aggregators: adaboostAggregator; arcx4Aggregator, vanillaAggregator, weightedAggregator; boost, boost.function, boost.list

Other arc-fs: arcfsReweighter

boostr documentation built on May 2, 2019, 1:42 p.m.