Package jhpro.nnet.jknnl

Kohonen neural networks are used in data mining proces and for knowledge discovery in databases.

See: Description

Interface Summary

Interface	Description
ActivationFunctionModel	Interface representing Neruon activation function model.
LearningDataModel	Object contains learning data.
LearningFactorFunctionalModel	Interface representing learning function model
MetricModel	Computes distance between pairs of objects.
NeighbourhoodFunctionModel	Neigbohood function interface
NetworkModel	Network Model interaface
NeuronModel	Neuron Model interface
TiredNeuronModel	Tired Neuon interface.
TopologyModel	Topology model interface

Class Summary

Class	Description
CityBlockMetric	City block metric return distance calculated by function: sum|x_i - y_i| for each element from inputs vectors, where x_i is first input vector element, y_i is second vector element.
ConstantFunctionalFactor	Constant Function factor.
Coords	Class used to get or set neuron coordinate in topology.
DefaultNetwork	Default Network Model.
DefaultNeuron	Class representing default neuron with specified activation function
DefaultNeuronWithBias	Class representing neuron with bias and specified activation function
EuclidesMetric	Euclid metric return distance calculated by function: sum[sqrt(x_i - y_i)] for each element from inputs vectors, where x_i is first input vector element, y_i is second vector element.
ExponentionalFunctionFactor	Exponential Function Factor.
GaussFunctionalFactor	Gauss Function Factor described by function exp(-(k^2))/(2*r^2))
GaussNeighbourhoodFunction	Gauss neighborhood Function
HardLimitActivationFunction	Hard Limit Activation Funciton return 1 if input value is greater then threshold and 0 otherwise.
HexagonalTopology	HexagonalTopology is a topology where neurons are placed in hexagonal net.
HiperbolicFunctionalFactor	Hyperbolic Function describe by function c1/(c2 + k) where: c1, c2 - parameters k - iteration number
KohonenNeuron	Class representing neuron with bias and specified activation function
KohonenNeuronWithTired	Class representing kohenen neuron with tiredness with specyfied activation function
LearningData	Object containing learning data.
LinearActivationFunction	Linear Activation Function return value calculated by function describe as y = a * x + b.
LinearFunctionalFactor	Linear function describe by: y = (n0 / maxIter) * (maxIter - iter) where: y - output value n0 - maximal factor maxIter - maximal number of iteration iter - iteration number
MatrixTopology	Matrix Topology is a topology where neurons is set in rows and columns.
MinkowskiMetric	Minkowski metric return distance calculated by function: {sum[(x_i - y_i)^p]}^(1/p)
SymetricHardLimitActivationFunction	Symmetric Hard Limit function return 1 if input value is greater then threshold and -1 otherwise.
TransparentActivationFunction	Transparent activation function return input value without any changes.
WTALearningFunction	Winner Takes All - algorithm there only wining neuron weights are changed according to the formula w(k+1) = w(k) + n * (x-w) where w(k+1) - neuron weight in k +1 iteration w(k) - neuron weight for k iteration n - value of learning function factor for k iteration x - learning vector of data w - neuron weight
WTALearningFunctionWithTired	WTALearningFunctionWithTired class - learning class used to learn neurons with tiredness
WTMLearningFunction	Winner Takes Most - algorithm where wining neuron and neurons in neighborhood weights are changed according to the formula w(k+1) = w(k) + n *N(i,x)* (x-w) where w(k+1) - neuron weight in k +1 iteration w(k) - neuron weight for k iteration n - value of learning function factor for k iteration N(i,x) - value of neighborhood function for i - specified neuron x - learning vector od data w - neuron weight
WTMLearningFunctionWithTired	WTMLearningFunctionWithTired class - learning class used to learn neuron with tiredness

Package jhpro.nnet.jknnl Description

Kohonen neural networks are used in data mining proces and for knowledge discovery in databases. Kohonen neural network library is a set of classes and functions used to design, train and calculates results from Kohonen neural network known as self organizing map.