Register xplus() as a parsnip model (pu_xplus with engine "xplus")
so PU learning composes with tidymodels: tune::tune_grid(),
finetune::tune_sim_anneal(), and rsample resamples all work against
the backend contract. The backend argument selects the learner — a
built-in name, an xplus_backend() spec, or a constructor name —
and backend-constructor arguments (trees, learn_rate, tree_depth,
min_n, mtry, sample_size, loss_reduction) are tunable through
tune() placeholders. Registration happens automatically when parsnip
is installed; call this function to re-register after loading parsnip
manually.
Examples
if (FALSE) { # \dontrun{
library(parsnip)
spec <- pu_xplus(backend = "xgboost", trees = 300, learn_rate = 0.05) |>
set_engine("xplus")
fit <- fit_xy(spec, x = data.frame(x), y = factor(y, levels = 0:1))
} # }