Create Omics-Optimized MLP Learner — make_mlp

Creates a Multi-Layer Perceptron with regularization optimized for high-dimensional omics data.

Usage

make_mlp_learner(
  n_hidden = 128L,
  n_layers = 2L,
  dropout = 0.5,
  batch_size = 32L,
  epochs = 100L,
  learning_rate = 0.001,
  weight_decay = 0.01,
  early_stopping_patience = 10L
)

Arguments

n_hidden: Number of hidden units per layer (default: 128)
n_layers: Number of hidden layers (default: 2)
dropout: Dropout rate (default: 0.5)
batch_size: Training batch size (default: 32)
epochs: Number of training epochs (default: 100)
learning_rate: Learning rate (default: 0.001)
weight_decay: L2 regularization (default: 0.01)
early_stopping_patience: Patience for early stopping (default: 10)

Value

An mlr3 Learner object

Details

The MLP is configured with: - High dropout (0.5) to prevent overfitting on high-dimensional data - Strong L2 regularization - Batch normalization between layers - Adam optimizer with early stopping

Examples

if (FALSE) { # \dontrun{
learner <- make_mlp_learner(n_hidden = 64, dropout = 0.5)
learner$train(task)
} # }