Fit Autoencoder — autoencoder_fit • OmicSelector

Fits a torch autoencoder and returns an OmicAutoencoder object.

Usage

autoencoder_fit(x, latent_dim = 32L, hidden_layers = c(128L, 64L),
  activation = c("relu", "tanh", "sigmoid", "gelu"), dropout = 0.1,
  epochs = 100L, batch_size = 32L, lr = 0.001, weight_decay = 0,
  loss = c("mse", "mae"), validation_split = 0.1,
  early_stopping = TRUE, patience = 10L, min_delta = 0,
  device = "cpu", seed = NULL, pretrained = NULL,
  freeze_encoder = FALSE)

Arguments

x: Numeric matrix or data.frame (samples x features).
latent_dim: Integer, size of latent representation.
hidden_layers: Integer vector of hidden layer sizes.
activation: Activation function.
dropout: Dropout rate between 0 and 1.
epochs: Number of training epochs.
batch_size: Batch size.
lr: Learning rate.
weight_decay: L2 weight decay.
loss: Loss function: "mse" or "mae".
validation_split: Fraction of data for validation.
early_stopping: Logical, enable early stopping.
patience: Early stopping patience.
min_delta: Minimum improvement to reset patience.
device: "cpu" or "cuda".
seed: Optional random seed.
pretrained: Optional path to saved autoencoder state.
freeze_encoder: Logical, freeze encoder weights during training.

Value

An OmicAutoencoder object.