tuning tweedie_variance_power parameter in xgboost

[chrisjacques13]

is there a way to tune the tweedie_variance_power parameter when using engine xgboost in tidymodels? It's not in the current list for boost_tree so this doesn't work:

xgb_model <- 
  boost_tree(
    trees = 1000, 
    tree_depth = tune(), 
    min_n = tune(), 
    loss_reduction = tune(),   
    sample_size = tune(), 
    mtry = tune(),       
    learn_rate = tune() ,
 tweedie_variance_power = tune()
  ) %>% 
  set_engine("xgboost", 
             nthreads = parallel::detectCores(),
             objective='reg:tweedie') %>% 
  set_mode("regression")

[tonyk7440]

I don't think that there is yet.

You could make a feature request in the dial repo issues

[Max]

You would move the power argument to set_engine() then manually make your own grid. See this blog post.

You could also put in an issue in the dials package to make automatic grid usage happen.