cross validation mrIMLpredicts for regression

[statslover123]

Hello

I am uncertain on how the k parameter works, your description here states k | A numeric sets the number of folds in the 10-fold cross-validation. 10 is the default.

Does that mean the default is 10 repetitions of 10-fold cross validation? (100 total test runs) or one repetition of 10 fold cross validation?

[nfj1380]

Hi there,

One repetition of 10 folds (k is just the number of folds). Sorry for the confusion - will fix this.

Cheers,

Nick

[statslover123]

that is awesome thank you, if I am tuning a nnet for regression, I can specify my hyperparameters in a grid and then use what number of folds I want to test with, using this command? thank you for the help

[nfj1380]

Yes that's right! On the topic of nnets - at least for classification problems 'brulee' MLP models now work https://brulee.tidymodels.org/reference/brulee_mlp.html i.e model_dnn <- mlp( hidden_units = 10, # 10 to start with dropout = tune(), epochs = tune(), learn_rate = tune(), activation = "elu" ) %>% set_engine("brulee") |> set_mode("classification")

May or may not be useful for regression problems.

[statslover123]

thank you! trying this now finally