mulan.classifier.neural.model
Class ActivationFunction

java.lang.Object
  mulan.classifier.neural.model.ActivationFunction

All Implemented Interfaces:

Serializable

Direct Known Subclasses:

ActivationLinear, ActivationTANH

public abstract class ActivationFunction
extends Object
implements Serializable

Abstract base class for activation functions. The activation function is used in neural network to transform an input of each layer (neuron) and produce the output for next layer (neuron). Depending on learning algorithm, derivation of activation function might be necessary.

Version:

2012.02.27

Author:

Jozef Vilcek

See Also:

Serialized Form

Constructor Summary

ActivationFunction()

Method Summary

abstract double	activate(double input) Computes an output value of the function for given input.
abstract double	derivative(double input) Computes an output value of function derivation for given input.
abstract double	getMax() Gets the maximum value the function can output.
abstract double	getMin() Gets the minimum value the function can output.

Methods inherited from class java.lang.Object

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

Constructor Detail

ActivationFunction

public ActivationFunction()

Method Detail

activate

public abstract double activate(double input)

Computes an output value of the function for given input.

Parameters:

input - the input value to the function

Returns:

the output value

derivative

public abstract double derivative(double input)

Computes an output value of function derivation for given input.

Parameters:

input - the input value to the function

Returns:

the output value

getMax

public abstract double getMax()

Gets the maximum value the function can output.

Returns:

maximum value of the function

getMin

public abstract double getMin()

Gets the minimum value the function can output.

Returns:

minimum value of the function