lightning_uq_box.models

class lightning_uq_box.models.MLP(dropout_p: float = 0.0, n_inputs: int = 1, n_hidden: list[int] = [100], n_outputs: int = 1, activation_fn: Module | None = None)

Multi-layer perceptron for predictions.

__init__(dropout_p: float = 0.0, n_inputs: int = 1, n_hidden: list[int] = [100], n_outputs: int = 1, activation_fn: Module | None = None) → None

Initialize a new instance of MLP.

Parameters:

• dropout_p – dropout percentage

• n_inputs – size of input dimension

• n_hidden – list of hidden layer sizes

• n_outputs – number of model outputs

• predict_sigma – whether the model intends to predict sigma term when minimizing NLL

• activation_fn – what nonlinearity to include in the network

forward(x) → Tensor

Forward pass through the neural network.

Parameters:

x – input vector to NN of dimension [batch_size, n_inputs]

Returs:

output from neural net of dimension [batch_size, n_outputs]