nerva_torch.loss_functions_torch

Thin wrappers around PyTorch loss modules for comparison and testing.

Functions

`negative_likelihood_loss_torch`(Y, T)	Computes the negative likelihood loss between Y and T.
`softmax_cross_entropy_loss_torch`(Y, T)	Computes the softmax cross entropy loss between Y and T.
`squared_error_loss_torch`(Y, T)	Computes the squared error loss between Y and T.

nerva_torch.loss_functions_torch.squared_error_loss_torch(Y, T)[source]

Computes the squared error loss between Y and T.

Parameters: Y (torch.Tensor): The predicted values. T (torch.Tensor): The target values.

Returns: float: The computed loss.

nerva_torch.loss_functions_torch.softmax_cross_entropy_loss_torch(Y, T)[source]

Computes the softmax cross entropy loss between Y and T.

Parameters: Y (torch.Tensor): The predicted values. T (torch.Tensor): The target values.

Returns: float: The computed loss.

nerva_torch.loss_functions_torch.negative_likelihood_loss_torch(Y, T)[source]

Computes the negative likelihood loss between Y and T. Note that PyTorch does not apply the log function, since it assumes Y is the output of a log softmax layer. For this reason we omit “log” in the name.

Parameters: Y (torch.Tensor): The predicted values. T (torch.Tensor): The target values.

Returns: float: The computed loss.