package jsat.lossfunctions;
/**
* The SquaredLoss loss function for regression L(x, y) = (x-y)2.
*
* This function is twice differentiable.
*
* @author Edward Raff
*/
public class SquaredLoss implements LossR
{
private static final long serialVersionUID = 130786305325167077L;
/**
* Computes the SquaredLoss loss
*
* @param pred the predicted value
* @param y the true value
* @return the squared loss
*/
public static double loss(double pred, double y)
{
final double x = y - pred;
return x * x * 0.5;
}
/**
* Computes the first derivative of the squared loss
*
* @param pred the predicted value
* @param y the true value
* @return the first derivative of the squared loss
*/
public static double deriv(double pred, double y)
{
return (pred - y);
}
/**
* Computes the second derivative of the squared loss, which is always
* {@code 1}
*
* @param pred the predicted value
* @param y the true value
* @return the second derivative of the squared loss
*/
public static double deriv2(double pred, double y)
{
return 1;
}
public static double regress(double score)
{
return score;
}
@Override
public double getLoss(double pred, double y)
{
return loss(pred, y);
}
@Override
public double getDeriv(double pred, double y)
{
return deriv(pred, y);
}
@Override
public double getDeriv2(double pred, double y)
{
return deriv2(pred, y);
}
@Override
public double getConjugate(double b, double pred, double y)
{
return b*b*0.5+b*y;
}
@Override
public double getDeriv2Max()
{
return 1;
}
@Override
public SquaredLoss clone()
{
return this;
}
@Override
public double getRegression(double score)
{
return score;
}
@Override
public double lipschitz()
{
//see Stochastic Dual Coordinate Ascent Methods for Regularized Loss Minimization
//"Both log loss and squared loss are 1-smooth"
return 1.0;
}
}
Ads help maintain this website.