jsat.classifiers.neuralnetwork.initializers

Class TanhInitializer

java.lang.Object

jsat.classifiers.neuralnetwork.initializers.TanhInitializer

All Implemented Interfaces:

java.io.Serializable, BiastInitializer, WeightInitializer

public class TanhInitializer
extends java.lang.Object
implements WeightInitializer, BiastInitializer

This initializer samples the weights from an adjusted uniform distribution in order to provided better behavior of neuron activation and gradients

See: Glorot, X., & Bengio, Y. (2010). Understanding the difficulty of training deep feedforward neural networks. Journal of Machine Learning Research - Proceedings Track, 9, 249–256. Retrieved from here

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
TanhInitializer()

Method Summary

Modifier and Type	Method and Description
TanhInitializer	clone()
void	init(Matrix w, java.util.Random rand) Initializes the values of the given weight matrix
void	init(Vec b, int fanIn, java.util.Random rand) Performs the initialization of the given vector of bias values

Methods inherited from class java.lang.Object

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

Constructor Detail

TanhInitializer

public TanhInitializer()

Method Detail

init

public void init(Matrix w,
                 java.util.Random rand)

Description copied from interface: WeightInitializer

Initializes the values of the given weight matrix

Specified by:

init in interface WeightInitializer

Parameters:

w - the matrix to initialize

rand - the source of randomness for the initialization

init

public void init(Vec b,
                 int fanIn,
                 java.util.Random rand)

Description copied from interface: BiastInitializer

Performs the initialization of the given vector of bias values

Specified by:

init in interface BiastInitializer

Parameters:

b - the vector to store the biases in

fanIn - the number of connections coming into the layer that these biases are for.

rand - the source of randomness for initialization

clone

public TanhInitializer clone()

Specified by:

clone in interface BiastInitializer

Specified by:

clone in interface WeightInitializer

Overrides:

clone in class java.lang.Object