fix: prevent defunct loss parameter values during regression tasks

PlantedML · Nov 17, 2023 · 9db1f77 · 9db1f77
1 parent 249090b
commit 9db1f77
Showing 1 changed file with 5 additions and 4 deletions.
diff --git a/R/rpf.R b/R/rpf.R
@@ -161,11 +161,12 @@ rpf_bridge <- function(processed, max_interaction = 1, ntrees = 50, splits = 30,
  checkmate::assert_number(delta, lower = 0, upper = 1)
  checkmate::assert_number(epsilon, lower = 0, upper = 1)
 
- # "median" is implemented but discarded
- checkmate::assert_choice(
- loss,
- choices = c("L1", "L2", "logit", "exponential")
+ # "median" loss is implemented but discarded
+ loss_functions <- switch(outcomes$mode,
+ "regression" = "L2",
+ "classification" = c("L1", "L2", "logit", "exponential")
  )
+ checkmate::assert_choice(loss, choices = loss_functions)
 
  checkmate::assert_flag(deterministic)
  checkmate::assert_int(nthreads, lower = 1L)