ANN

Simple Dense Model

class bias_amplification.attacker_models.ANN.simpleDenseModel(input_dims, output_dims, num_layers=5, numFirst=16, activations=['relu', 'relu', 'sigmoid', 'relu', 'relu'])[source]

Bases: Module

This class defines a simple dense neural network model, used mainly for the attacker models in the bias amplification analysis.

Methods

`count_params`()	This function counts the number of parameters in the neural network.
`filterActivations`()	This function filters the activations to be used in the neural network.
`forward`(x)	This function performs the forward pass through the neural network.
`initLayers`()	This function initialises the layers of the neural network.

__init__(input_dims, output_dims, num_layers=5, numFirst=16, activations=['relu', 'relu', 'sigmoid', 'relu', 'relu'])[source]

This function initialises the neural network.

Parameters:

input_dims: int: The number of input dimensions.
output_dims: int: The number of output dimensions.
num_layers: int: The number of layers in the neural network.
numFirst: int: The number of neurons in the first layer.
activations: list: The activations to use in the layers.

count_params()[source]

This function counts the number of parameters in the neural network.

Returns:

int: The number of parameters in the neural network.

filterActivations()[source]

This function filters the activations to be used in the neural network. If the number of activations is less than the number of layers, the activations are extended with the identity activation.

Parameters:

activations: list: The activations to use in the layers.

Returns:

None

forward(x)[source]

This function performs the forward pass through the neural network.

Parameters:

x: torch.Tensor: The input to the neural network.

Returns:

x: torch.Tensor: The output of the neural network.

initLayers()[source]

This function initialises the layers of the neural network.

Parameters:

input_dims: int: The number of input dimensions.
output_dims: int: The number of output dimensions.
num_layers: int: The number of layers in the neural network.
numFirst: int: The number of neurons in the first layer.
activations: list: The activations to use in the layers.

Returns:

None

Raises:

ValueError: If the number of layers is less than 1.

training: bool