Base learner: Multilayer perceptron with brulee

Multilayer perceptron model with different configurations of hidden units, dropout, activation, and learning rate using brulee and tidymodels. With proper settings, users can utilize graphics processing units (GPU) to speed up the training process.

Usage

fit_base_brulee(
  dt_imputed,
  folds = NULL,
  tune_mode = "grid",
  tune_bayes_iter = 50L,
  learn_rate = 0.1,
  yvar = "Arithmetic.Mean",
  xvar = seq(5, ncol(dt_imputed)),
  vfold = 5L,
  device = "cuda:0",
  trim_resamples = TRUE,
  return_best = FALSE,
  ...
)

Arguments

dt_imputed

The input data table to be used for fitting.

folds

pre-generated rset object with minimal number of columns. If NULL, vfold should be numeric to be used in rsample::vfold_cv.

tune_mode

character(1). Hyperparameter tuning mode. Default is "grid", "bayes" is acceptable.

tune_bayes_iter

integer(1). The number of iterations for Bayesian optimization. Default is 50. Only used when tune_mode = "bayes".

learn_rate

The learning rate for the model. For branching purpose. Default is 0.1.

yvar

The target variable.

xvar

The predictor variables.

vfold

The number of folds for cross-validation.

device

The device to be used for training. Default is "cuda:0". Make sure that your system is equipped with CUDA-enabled graphical processing units.

trim_resamples

logical(1). Default is TRUE, which replaces the actual data.frames in splits column of tune_results object with NA.

return_best

logical(1). If TRUE, the best tuned model is returned.

...

Additional arguments to be passed.

Value

The fitted workflow.

Details

Hyperparameters hidden_units, dropout, activation, and learn_rate are tuned. With tune_mode = "grid", users can modify learn_rate explicitly, and other hyperparameters will be predefined (56 combinations per learn_rate).

Note

tune package should be 1.2.0 or higher. brulee should be installed with GPU support.