boone.map

Class Function.SoftMax

java.lang.Object

boone.map.Function

boone.map.Function.SoftMax

All Implemented Interfaces:

Storable, java.lang.Cloneable

Enclosing class:

Function

public static class Function.SoftMax
extends Function

The softmax function mapping an array of values to a probability distribution by f(xi) = e^(xi) / sumi(e^(xi)). Note that the mapping function is not the standard Function.map().

Nested Class Summary

Nested classes/interfaces inherited from class boone.map.Function

Function.AboutEqual, Function.AtLeast, Function.AtMost, Function.CatmullRomSpline, Function.Clip, Function.Composition, Function.Exponential, Function.GreaterThan, Function.Identity, Function.LessThan, Function.NaturalSpline, Function.NotImplemented, Function.ReLU, Function.Scaled, Function.Sigmoid, Function.Signum, Function.SignumWithoutZero, Function.Sinus, Function.SoftMax, Function.Spline, Function.TanH

Constructor Summary

Constructors

Constructor and Description
SoftMax()

Method Summary

Modifier and Type	Method and Description
double	getMaxValue() Return the largest possible function value.
double	getMinValue() Return the smallest possible function value.
double	map(double x) Maps values to a probability distribution.
void	setParameters(java.lang.Object params) Set parameters, if needed.

Methods inherited from class boone.map.Function

clone, create, fromXML, mapDerivative, mapSecondDerivative, toXML

Methods inherited from class java.lang.Object

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

Constructor Detail

SoftMax

public SoftMax()

Method Detail

setParameters

public void setParameters(java.lang.Object params)

Description copied from class: Function

Set parameters, if needed.

Overrides:

setParameters in class Function

Parameters:

params - the parameters

map

public double map(double x)

Maps values to a probability distribution. Computes a stable softmax by subtracting the maximal value from the arguments so that the nominator becomes 1.0 and the denominator is limited to a safe range. The given x-value is not a real argument. If it is less than 0, it indicates that all values in the parameter array 'values' are mapped. If it is positive, then it is an index to a single value in 'values' transformed by softMax.

Overrides:

map in class Function

Parameters:

x - a (negative) index into values

getMinValue

public double getMinValue()

Description copied from class: Function

Return the smallest possible function value. If no exact value can be produced easily, this should return an estimation (lower bound).

Specified by:

getMinValue in class Function

Returns:

the minimum function result.

getMaxValue

public double getMaxValue()

Description copied from class: Function

Return the largest possible function value. If no exact value can be produced easily, this should return an estimation (upper bound).

Specified by:

getMaxValue in class Function

Returns:

the maximum function result.