Loss Functions

class gwu_nn.loss_functions.LossFunction

Abstract Class for loss functions

abstract loss(y_true, y_pred)

Calculates the loss for the given network.

Args:

y_true (np.array): the true values y_pred (np.array): the predicted values

Returns:

np.array(np.array): the network’s loss

abstract loss_partial_derivative(y_true, y_pred)

Calculates the derivative of the loss for the given network.

Args:

y_true (np.array): the true values y_pred (np.array): the predicted values

Returns:

np.array(np.array): the partial derivative for the network’s loss

class gwu_nn.loss_functions.MSE

Class for implementing the MSE loss function. Inheirits loss and loss_partial_derivative from LossFunction

classmethod loss(y_true, y_pred)

Calculates the MSE for the true vs predicted values

Returns:

np.array: MSE for each input

classmethod loss_partial_derivative(y_true, y_pred)

Calculates the derivative of the MSE for each prediction

Returns:

np.array: Partial derivative of the MSE

class gwu_nn.loss_functions.LogLoss

Class for implementing the LogLoss loss function. Inheirits loss and loss_partial_derivative from LossFunction

classmethod loss(y_true, y_pred)

Calculates the LogLoss for the true vs predicted values

Returns:

np.array: LogLoss for each input

classmethod loss_partial_derivative(y_true, y_pred)

Calculates the derivative of the LogLoss for each prediction

Returns:

np.array: Partial derivative of the LogLoss

class gwu_nn.loss_functions.CrossEntropy

Class for implementing the CrossEntropy loss function. Inheirits loss and loss_partial_derivative from LossFunction

classmethod loss(y_true, y_pred)

Calculates the CrossEntropy for the true vs predicted classes

Returns:

np.array: CrossEntropy for each input/class

classmethod loss_partial_derivative(y_true, y_pred)

Calculates the derivative of the CrossEntropy for each prediction

Returns:

np.array: Partial derivative of the CrossEntropy