Package com.jml.neural_network.activations

Class Softmax

java.lang.Object

com.jml.neural_network.activations.Softmax

All Implemented Interfaces:

ActivationFunction

public class Softmax
extends Object
implements ActivationFunction

The softmax activation function. f(x)_i = exp(x_i) / sum_j=1^m( exp(x_j) ) where x is a vector of length m and sum_j=1^m ( exp(x_j) ) = exp(x₁) + exp(x₂) + ... + exp(x_m).

Field Summary

Fields

Modifier and Type	Field	Description
String	NAME

Constructor Summary

Constructors

Constructor	Description
Softmax()

Method Summary

Modifier and Type	Method	Description
linalg.Matrix	back(linalg.Matrix data)	Applies the derivative of the activation function, element-wise, to a matrix.
linalg.Matrix	forward(linalg.Matrix data)	Applies the activation function, element-wise, to a matrix.
String	getName()	Gets the name of the activation function.

Methods inherited from class java.lang.Object

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

Field Details

NAME

public final String NAME

See Also:

Constant Field Values

Constructor Details

Softmax

public Softmax()

Method Details

forward

public linalg.Matrix forward(linalg.Matrix data)

Applies the activation function, element-wise, to a matrix.

Specified by:

forward in interface ActivationFunction

Parameters:

data - The matrix to apply activation function to.

Returns:

The result of the element-wise activation function applied to the matrix.

back

public linalg.Matrix back(linalg.Matrix data)

Applies the derivative of the activation function, element-wise, to a matrix.

Specified by:

back in interface ActivationFunction

Parameters:

data - The matrix to apply the derivative of the activation function to.

Returns:

The slope of the activation function, evaluated element-wise, of the data matrix.

getName

public String getName()

Gets the name of the activation function.

Specified by:

getName in interface ActivationFunction

Returns:

The name of the activation function as a String.