Offline Surrogate-Performance Table

The scoring core of replay_surrogate_performance(): refits learner on every archive batch prefix, scores it on the held-out task, and returns the per-batch performance table with one row per batch and one column per measure – the same shape CallbackSurrogatePerformance records online (columns batch_nr, n_evals, then the measure columns). Use this when the full multi-measure table is wanted rather than a single-target TaskLCE.

Usage

replay_surrogate_perf_table(
  archive,
  learner,
  task,
  measures = list(msr("regr.rsq"), msr("regr.mae"))
)

Arguments

archive

(bbotk::ArchiveBatch)
Archive of a finished single-target run.

learner

(mlr3::LearnerRegr | mlr3mbo::SurrogateLearner)
Surrogate model refit per batch; see replay_surrogate_performance().

task

(mlr3::TaskRegr)
Held-out regression task to score on.

measures

(list() of mlr3::Measure)
Regression measures; named entries use their names as column names.

Value

data.table::data.table with columns batch_nr, n_evals, and one numeric column per measure.