|
||||||||||
PREV CLASS NEXT CLASS | FRAMES NO FRAMES | |||||||||
SUMMARY: NESTED | FIELD | CONSTR | METHOD | DETAIL: FIELD | CONSTR | METHOD |
java.lang.Objectorg.encog.ml.BasicML
org.encog.neural.art.ART
org.encog.neural.art.ART1
public class ART1
Implements an ART1 neural network. An ART1 neural network is trained to recognize bipolar patterns as it is presented data. There is no distinct learning phase, like there is with other neural network types. The ART1 neural network is a type of Adaptive Resonance Theory (ART) neural network. ART1 was developed by Stephen Grossberg and Gail Carpenter. This neural network type supports only bipolar input. The ART1 neural network is trained as it is used. New patterns are presented to the ART1 network, and they are classified into either new, or existing, classes. Once the maximum number of classes have been used the network will report that it is out of classes. ART1 neural networks are used for classification. There are essentially 2 layers in an ART1 network. The first, named the F1 layer, acts as the input. The F1 layer receives bipolar patterns that the network is to classify. The F2 layer specifies the maximum number classes that the ART1 network can recognize. Plasticity is an important part for all Adaptive Resonance Theory (ART) neural networks. Unlike most neural networks, ART1 does not have a distinct training and usage stage. The ART1 network will learn as it is used.
Field Summary |
---|
Fields inherited from class org.encog.neural.art.ART |
---|
PROPERTY_A1, PROPERTY_B1, PROPERTY_C1, PROPERTY_D1, PROPERTY_L, PROPERTY_NO_WINNER, PROPERTY_VIGILANCE |
Constructor Summary | |
---|---|
ART1()
Default constructor, used mainly for persistence. |
|
ART1(int theF1Count,
int theF2Count)
Construct the ART1 network. |
Method Summary | |
---|---|
void |
adjustWeights()
Adjust the weights for the pattern just presented. |
int |
classify(MLData input)
Classify the input data to a class number. |
void |
compute(BiPolarNeuralData input,
BiPolarNeuralData output)
Compute the output from the ART1 network. |
MLData |
compute(MLData input)
Compute the output for the BasicNetwork class. |
double |
getA1()
|
double |
getB1()
|
double |
getC1()
|
double |
getD1()
|
int |
getF1Count()
|
int |
getF2Count()
|
int |
getInputCount()
|
double |
getL()
|
int |
getNoWinner()
|
int |
getOutputCount()
|
double |
getVigilance()
|
Matrix |
getWeightsF1toF2()
|
Matrix |
getWeightsF2toF1()
|
int |
getWinner()
|
boolean |
hasWinner()
|
double |
magnitude(BiPolarNeuralData input)
Get the magnitude of the specified input. |
void |
reset()
Reset the weight matrix back to starting values. |
void |
reset(int seed)
Reset with a specic seed. |
void |
setA1(double theA1)
Set the A1 parameter. |
void |
setB1(double theB1)
Set the B1 parameter. |
void |
setC1(double theC1)
Set the C1 parameter. |
void |
setD1(double theD1)
Set the D1 parameter. |
void |
setF1Count(int i)
Set the F1 count. |
void |
setF2Count(int i)
Set the F2 count. |
void |
setL(double theL)
Set the L parameter. |
void |
setNoWinner(int i)
Set the i parameter. |
void |
setVigilance(double theVigilance)
Set the vigilance. |
void |
setWeightsF1toF2(Matrix matrix)
Set the f1 to f2 matrix. |
void |
setWeightsF2toF1(Matrix matrix)
Set the f2 to f1 matrix. |
Methods inherited from class org.encog.neural.art.ART |
---|
updateProperties |
Methods inherited from class org.encog.ml.BasicML |
---|
getProperties, getPropertyDouble, getPropertyLong, getPropertyString, setProperty, setProperty, setProperty |
Methods inherited from class java.lang.Object |
---|
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
Constructor Detail |
---|
public ART1()
public ART1(int theF1Count, int theF2Count)
theF1Count
- The neuron count for the f1 layer.theF2Count
- The neuron count for the f2 layer.Method Detail |
---|
public void adjustWeights()
public int classify(MLData input)
classify
in interface MLClassification
input
- The input data.
public void compute(BiPolarNeuralData input, BiPolarNeuralData output)
input
- The input to the network.output
- The output from the network.public MLData compute(MLData input)
input
- The input to the network.
public double getA1()
public double getB1()
public double getC1()
public double getD1()
public int getF1Count()
public int getF2Count()
public int getInputCount()
getInputCount
in interface MLInput
public double getL()
public int getNoWinner()
public int getOutputCount()
getOutputCount
in interface MLOutput
public double getVigilance()
public Matrix getWeightsF1toF2()
public Matrix getWeightsF2toF1()
public int getWinner()
public boolean hasWinner()
public double magnitude(BiPolarNeuralData input)
input
- The input to calculate the magnitude for.
public void reset()
reset
in interface MLResettable
public void reset(int seed)
reset
in interface MLResettable
seed
- The seed to reset with.public void setA1(double theA1)
theA1
- The new value.public void setB1(double theB1)
theB1
- The new value.public void setC1(double theC1)
theC1
- The new value.public void setD1(double theD1)
theD1
- The new value.public void setF1Count(int i)
i
- The count.public void setF2Count(int i)
i
- The count.public void setL(double theL)
theL
- The new value.public void setNoWinner(int i)
i
- The new value.public void setVigilance(double theVigilance)
theVigilance
- The new value.public void setWeightsF1toF2(Matrix matrix)
matrix
- The new matrix.public void setWeightsF2toF1(Matrix matrix)
matrix
- The new matrix.
|
||||||||||
PREV CLASS NEXT CLASS | FRAMES NO FRAMES | |||||||||
SUMMARY: NESTED | FIELD | CONSTR | METHOD | DETAIL: FIELD | CONSTR | METHOD |