Package org.ddogleg.optimization.loss

Class LossHuberSmooth

java.lang.Object

org.ddogleg.optimization.loss.LossFunctionBase

org.ddogleg.optimization.loss.LossHuberSmooth

Direct Known Subclasses:

LossHuberSmooth.Function, LossHuberSmooth.Gradient

public abstract class LossHuberSmooth
extends LossFunctionBase

Smooth approximation to the huber loss [1]. This is similar to the L1 Loss in Ceres.

L(a) = t2(sqrt(1+(a/t)2-1)

, where 'a' is the residual, and 't' is the passed in tuning. For small values it will approximate a²/2, but for large values it will be a line with slope 't'. The point of inflection where the functions begins to behave more linear is for values of 'a' > 't'.

1. Huber Loss - Wikipedia 2023

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	LossHuberSmooth.Function	Implementation of the smooth Huber loss function
static class	LossHuberSmooth.Gradient	Implementation of the smooth Huber loss gradient

Field Summary

Fields inherited from class org.ddogleg.optimization.loss.LossFunctionBase

numberOfFunctions

Constructor Summary

Constructors

Modifier	Constructor	Description
protected	LossHuberSmooth(double threshold)

Method Summary

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

LossHuberSmooth

protected LossHuberSmooth(double threshold)