gossipy.model.nn module — gossipy 0.0.1 documentation

class gossipy.model.nn.AdaLine(dim)#

Bases: gossipy.model.TorchModel

The Adaline perceptron model.

Implementation of the AdaLine perceptron model [], []. The model is a simple perceptron with a linear activation function.

Parameters: dim (int) – The number of input features.

forward(x)#

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Return type: Tensor
Parameters: x (torch.Tensor) –

get_size()#

Returns the number of parameters of the model.

Returns: The number of parameters of the model.
Return type: int

init_weights()#

Initialize the weights of the model.

Return type: None

class gossipy.model.nn.LogisticRegression(input_dim, output_dim)#

Bases: gossipy.model.TorchModel

Logistic regression model.

Implementation of the logistic regression model.

Parameters

input_dim (int) – The number of input features.
output_dim (int) – The number of output neurons.

forward(x)#

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Return type: Tensor
Parameters: x (torch.Tensor) –

init_weights()#

Initialize the weights of the model.

Return type: None

class gossipy.model.nn.TorchMLP(input_dim, output_dim, hidden_dims=(100, ), activation=<class 'torch.nn.modules.activation.ReLU'>)#

Bases: gossipy.model.TorchModel

Multi-layer perceptron model.

Implementation of the multi-layer perceptron model. The model is composed of a sequence of linear layers with the specified activation function (same activation for all layers but the last one).

Parameters

input_dim (int) – The number of input features.
output_dim (int) – The number of output neurons.
hidden_dims (Tuple[int], default=(100,)) – The number of hidden neurons in each hidden layer.
activation (torch.nn.modules.activation, default=ReLU) – The activation function of the hidden layers.

forward(x)#

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Return type: Tensor
Parameters: x (torch.Tensor) –

init_weights()#

Initialize the weights of the model.

Return type: None

class gossipy.model.nn.TorchPerceptron(dim, activation=<class 'torch.nn.modules.activation.Sigmoid'>, bias=True)#

Bases: gossipy.model.TorchModel

Perceptron model.

Implementation of the perceptron model by Rosenblatt [].

Parameters

dim (int) – The number of input features.
activation (torch.nn.modules.activation, default=Sigmoid()) – The activation function of the output neuron.
bias (bool, optional) – Whether to add a bias term to the output neuron.

forward(x)#

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Return type: Tensor
Parameters: x (torch.Tensor) –

init_weights()#

Initialize the weights of the model.

Return type: None

gossipy.model.nn module#

Module contents#