Package de.htwdd.rosenkoenig.neuro.net

The package net contains implementations, examples and utilities to create and train neural networks.

See:
Description

Interface Summary

ActivationFunction

Interface for functions.

Class Summary

Cone	The famous cone shaped (sawtooth, like a Russian L) neighborhood function.
Cosine	Cosine shaped neighborhood function.
FeedForwardNet	This class implements a feed forward net.
Gauss	The Gauss neighborhood function.
Identity	This is a dummy function that returns the same value for compute that was given to the function and 1.0 for computeDerivation.
InnerLayer	This class represents all hidden layers.
InputLayer	This class represents an input layer.
KohonenLayer	This wrapper class is needed to use a KohonenNet as input layer for a FeedForwardNet.
KohonenNet	KohonenNet is an implementation of a self organizing map.
Layer	This abstract class is the base class for the layers used in this project.
NeighbourhoodFunction	Base class to provide a KohonenNet with a neighbourhood function.
NeuralNet	This is the abstract class that every neuronal net is supposed to derive from.
Sigmoid	A sigmoid activation function.
Synapse	This class is used to interconnect layers of neurons.
Tanh	An activation function using tanh.
WinnerTakesAll	The WinnerTakesAll neighbourhood function gives the same activation (1.0) to all neurons within the neighbourhood radius.

Package de.htwdd.rosenkoenig.neuro.net Description

The package net contains implementations, examples and utilities to create and train neural networks.
At the moment only implementations of feed forward networks and Kohonen networks have been implemented.