jsat.classifiers.neuralnetwork.initializers

Interface WeightInitializer

All Superinterfaces:

java.io.Serializable

All Known Implementing Classes:

GaussianNormalInit, TanhInitializer

public interface WeightInitializer
extends java.io.Serializable

This interface specifies the method of initializing the weight connections in a neural network.

Method Summary

Modifier and Type	Method and Description
WeightInitializer	clone()
void	init(Matrix w, java.util.Random rand) Initializes the values of the given weight matrix

Method Detail

init

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

Initializes the values of the given weight matrix

Parameters:

w - the matrix to initialize

rand - the source of randomness for the initialization

clone

WeightInitializer clone()