org.apache.reef.examples.group.bgd.loss

Interface LossFunction

All Known Implementing Classes:

LogisticLossFunction, SquaredErrorLossFunction, WeightedLogisticLossFunction

@DefaultImplementation(value=SquaredErrorLossFunction.class)
public interface LossFunction

Interface for Loss Functions.

Method Summary

Methods

Modifier and Type	Method and Description
double	computeGradient(double y, double f) Computes the gradient with respect to f, if y is the true label.
double	computeLoss(double y, double f) Computes the loss incurred by predicting f, if y is the true label.

Method Detail

computeLoss

double computeLoss(double y,
                 double f)

Computes the loss incurred by predicting f, if y is the true label.

Parameters:

y - the label

f - the prediction

Returns:

the loss incurred by predicting f, if y is the true label.

computeGradient

double computeGradient(double y,
                     double f)

Computes the gradient with respect to f, if y is the true label.

Parameters:

y - the label

f - the prediction

Returns:

the gradient with respect to f