void Initialize()
{
performanceCounter = new ProcessPerformanceCounter();
// Create fuzzy art map network.
network = new SimplifiedFuzzyARTMap(Parameter.InputSize * 2,
Parameter.NoOfClasses * 2, Parameter.ArtABasedVigilance, Parameter.ChoicingParam, Parameter.ArtABeta);
network.FastCommitedSlowLearningOption = true;
// Analyze max-min of input columns
AnalyzeMaxMin();
// Initialize inputtransformers level 1
inputTransformers1 = new DataTransformChain[Parameter.InputSize];
InitializeInputDataTransform(inputTransformers1, Parameter.InputSize);
// inputtransformation level 2
inputTransformers2 = new DataTransformChain();
inputTransformers2.AppendDataTransform(new ComplementDataTransform(1.0));
// Target column
targetTransform = new DataTransformChain();
targetTransform.AppendDataTransform(new OneOfNDataTransform(Parameter.NoOfClasses));
targetTransform.AppendDataTransform(new ComplementDataTransform(1.0));
// Output transform
outputTransformers = new DataTransformChain();
outputTransformers.AppendDataTransform(new UnComplementDataTransform());
outputTransformers.AppendDataTransform(new ReverseOneOfNDataTransform(Parameter.NoOfClasses));
}
public void Serialize(string filepath , SimplifiedFuzzyARTMap network)
{
Dictionary<String, String> props = new Dictionary<String, String>();
props.Add("InputSizeA", network.ArtA.InputSize.ToString());
props.Add("InputSizeB", network.ArtB.InputSize.ToString());
props.Add("ArtAVigilance", network.BasedVigilanceA.ToString());
props.Add("ArtAChoiceValue", network.ArtA.Choice.ToString());
props.Add("ArtABeta", network.ArtA.Beta.ToString());
// Write weights of neuron in f2 layer of ART-A
props.Add("ArtANeuronSize", network.ArtA.F2.Count.ToString());
for (int i = 1; i <= network.ArtA.F2.Count; i++)
{
StringBuilder sb = new StringBuilder();
for(int j = 0; j < network.ArtA.InputSize - 1; j++)
sb.Append(network.ArtA.F2.Neurons(i).Weights[j].ToString() + ",");
sb.Append(network.ArtA.F2.Neurons(i).Weights[network.ArtA.InputSize-1].ToString());
props.Add("ArtA-N-" + i.ToString(), sb.ToString());
}
// Write weights of neuron in f2 layer of ART-B
props.Add("ArtBNeuronSize", network.ArtB.F2.Count.ToString());
for (int i = 1; i <= network.ArtB.F2.Count; i++)
{
StringBuilder sb = new StringBuilder();
for(int j = 0; j < network.ArtB.InputSize - 1; j++)
sb.Append(network.ArtB.F2.Neurons(i).Weights[j].ToString() + ",");
sb.Append(network.ArtB.F2.Neurons(i).Weights[network.ArtB.InputSize-1].ToString());
props.Add("ArtB-N-" + i.ToString(), sb.ToString());
}
// Write weights of neuron in map field layer
props.Add("MapNeuronSize", network.Map.Count.ToString());
for (int i = 1; i <= network.Map.Count; i++)
{
StringBuilder sb = new StringBuilder();
for(int j = 0; j < network.ArtB.InputSize - 1; j++)
sb.Append(network.Map.Neurons(i).Weights[j].ToString() + ",");
sb.Append(network.Map.Neurons(i).Weights[network.ArtB.InputSize-1].ToString());
props.Add("Map-N-" + i.ToString(), sb.ToString());
}
StringBuilder data = new StringBuilder();
foreach (string key in props.Keys)
{
data.Append(key + "=" + props[key] + "\n");
}
StreamWriter writer = new StreamWriter(filepath);
writer.Write(data.ToString());
writer.Close();
}
void Initialize()
{
performanceCounter = new ProcessPerformanceCounter();
if (param.Mode == ProcessModifiedHybridFuzzyARTMapMode.INTERNAL ||
param.Mode == ProcessModifiedHybridFuzzyARTMapMode.DUAL)
{
mfartmap = new ModifiedFuzzyARTMap(param.InputSize, param.NoOfClasses, param.ChoicingParam,
param.ArtABasedVigilance, param.MaximumEntropy, param.MaximumTotalEntropy, param.VigilanceAdjustRate, param.ArtABeta);
}
else
{
fartmap = new SimplifiedFuzzyARTMap(2 * param.InputSize, 2 * param.NoOfClasses, param.ArtABasedVigilance,
param.ChoicingParam, param.ArtABeta);
}
if (param.Mode == ProcessModifiedHybridFuzzyARTMapMode.EXTERNAL ||
param.Mode == ProcessModifiedHybridFuzzyARTMapMode.DUAL)
{
mcoa = new ModifiedClusteringOrderingAlgorithms(param.NoOfClasses);
}
// Analyze max-min of input columns
AnalyzeMaxMin();
// Initialize inputtransformers level 1
inputTransformers1 = new DataTransformChain[Parameter.InputSize];
InitializeInputDataTransform(inputTransformers1, Parameter.InputSize);
// inputtransformation level 2
inputTransformers2 = new DataTransformChain();
inputTransformers2.AppendDataTransform(new ComplementDataTransform(1.0));
// Target column
targetTransform = new DataTransformChain();
targetTransform.AppendDataTransform(new OneOfNDataTransform(Parameter.NoOfClasses));
targetTransform.AppendDataTransform(new ComplementDataTransform(1.0));
// Output transform
outputTransformers = new DataTransformChain();
outputTransformers.AppendDataTransform(new UnComplementDataTransform());
outputTransformers.AppendDataTransform(new ReverseOneOfNDataTransform(Parameter.NoOfClasses));
}
public OrderedFuzzyARTMap(SimplifiedFuzzyARTMap fuzzyARTMap,
IOrderingAlgorithms orderingAlgorithms)
{
this.fuzzyARTMap = fuzzyARTMap;
this.orderingAlgorithms = orderingAlgorithms;
}
public OrderedFuzzyARTMap(SimplifiedFuzzyARTMap fuzzyARTMap, int noOfClusters) :
this(fuzzyARTMap, new MinMaxClusteringOrderingAlgorithms(noOfClusters))
{
}
void Initialize()
{
performanceCounter = new ProcessPerformanceCounter();
// Create fuzzy art map network.
network = new SimplifiedFuzzyARTMap(Parameter.InputSize * 2,
Parameter.NoOfClasses * 2, Parameter.ArtABasedVigilance, Parameter.ChoicingParam, Parameter.ArtABeta);
network.FastCommitedSlowLearningOption = true;
// Analyze max-min of input columns
AnalyzeMaxMin();
// Initialize inputtransformers level 1
inputTransformers1 = new DataTransformChain[Parameter.InputSize];
InitializeInputDataTransform(inputTransformers1, Parameter.InputSize);
// inputtransformation level 2
inputTransformers2 = new DataTransformChain();
inputTransformers2.AppendDataTransform(new ComplementDataTransform(1.0));
// Target column
targetTransform = new DataTransformChain();
targetTransform.AppendDataTransform(new OneOfNDataTransform(Parameter.NoOfClasses));
targetTransform.AppendDataTransform(new ComplementDataTransform(1.0));
// Output transform
outputTransformers = new DataTransformChain();
outputTransformers.AppendDataTransform(new UnComplementDataTransform());
outputTransformers.AppendDataTransform(new ReverseOneOfNDataTransform(Parameter.NoOfClasses));
}
public OrderedFuzzyARTMap(SimplifiedFuzzyARTMap fuzzyARTMap,
IOrderingAlgorithms orderingAlgorithms)
{
this.fuzzyARTMap = fuzzyARTMap;
this.orderingAlgorithms = orderingAlgorithms;
}
public OrderedFuzzyARTMap(SimplifiedFuzzyARTMap fuzzyARTMap, int noOfClusters)
: this(fuzzyARTMap, new MinMaxClusteringOrderingAlgorithms(noOfClusters))
{
}
void UpdateTrainingResult(SimplifiedFuzzyARTMap network, int epochs)
{
txtNoOfEpochs.Text = epochs.ToString();
txtNoOfNeuronARTA.Text = network.ArtA.F2.Count.ToString();
txtNoOfNeuronARTB.Text = network.ArtB.F2.Count.ToString();
txtNoOfNeuronMAPField.Text = network.Map.Count.ToString();
btnSaveWeights.Enabled = true;
lnkMoreDetails.Visible = true;
}
public SimplifiedFuzzyARTMap DeSerialize(string filepath)
{
StreamReader reader = new StreamReader(filepath);
string data = reader.ReadToEnd();
reader.Close();
SimplifiedFuzzyARTMap network = null;
if (data != null && data != String.Empty)
{
Dictionary<string, string> props = new Dictionary<string, string>();
// Insert key and value from data string.
string[] lines = data.Split('\n');
foreach (string line in lines)
{
string[] sections = line.Split('=');
props.Add(sections[0], sections[1]);
}
int inputSizeA = Convert.ToInt32(props["InputSizeA"]);
int inputSizeB = Convert.ToInt32(props["InputSizeB"]);
double artABasedVigilance = Convert.ToDouble(props["ArtAVigilance"]);
double artAChoiceValue = Convert.ToDouble(props["ArtAChoiceValue"]);
double artABeta = Convert.ToDouble(props["ArtABeta"]);
network = new SimplifiedFuzzyARTMap(inputSizeA, inputSizeB,
artABasedVigilance, artAChoiceValue, artABeta);
// Store weights of each neuron in art-a.
int artAF2NeuronSize = Convert.ToInt32(props["ArtANeuronSize"]);
for(int i = 1; i <= artAF2NeuronSize; i++)
{
string weightsStr = props["ArtA-N-" + i];
Neuron neuron = new Neuron(LinearFunction.Instance);
string[] weights = weightsStr.Split(',');
for (int j = 0; j < inputSizeA; j++)
neuron.Weights.Add(Convert.ToDouble(weights[j]));
network.ArtA.F2.AddNeuron(neuron);
}
// Store weights of each neuron in art-b.
int artBF2NeuronSize = Convert.ToInt32(props["ArtBNeuronSize"]);
for(int i = 1; i <= artBF2NeuronSize; i++)
{
string weightsStr = props["ArtB-N-" + i];
Neuron neuron = new Neuron(LinearFunction.Instance);
string[] weights = weightsStr.Split(',');
for (int j = 0; j < inputSizeB; j++)
neuron.Weights.Add(Convert.ToDouble(weights[j]));
network.ArtB.F2.AddNeuron(neuron);
}
// Store weights of each neuron in map field layer.
int mapNeuronSize = Convert.ToInt32(props["MapNeuronSize"]);
for (int i = 1; i <= mapNeuronSize; i++)
{
string weightsStr = props["Map-N-" + i];
Neuron neuron = new Neuron(LinearFunction.Instance);
string[] weights = weightsStr.Split(',');
for (int j = 0; j < inputSizeB; j++)
neuron.Weights.Add(Convert.ToDouble(weights[j]));
network.Map.AddNeuron(neuron);
}
}
return network;
}
public void TearDown()
{
fuzzy = null;
}
public void SetUp()
{
fuzzy = new SimplifiedFuzzyARTMap(INPUT_SIZE, TARGET_SIZE,
ART_A_BASED_VIGILANCE, ART_A_CHOICE_VALUE, ART_A_BETA);
}
void Initialize()
{
performanceCounter = new ProcessPerformanceCounter();
if (param.Mode == ProcessModifiedHybridFuzzyARTMapMode.INTERNAL ||
param.Mode == ProcessModifiedHybridFuzzyARTMapMode.DUAL)
{
mfartmap = new ModifiedFuzzyARTMap(param.InputSize, param.NoOfClasses, param.ChoicingParam,
param.ArtABasedVigilance, param.MaximumEntropy, param.MaximumTotalEntropy, param.VigilanceAdjustRate, param.ArtABeta);
}
else
{
fartmap = new SimplifiedFuzzyARTMap(2*param.InputSize, 2*param.NoOfClasses, param.ArtABasedVigilance,
param.ChoicingParam, param.ArtABeta);
}
if (param.Mode == ProcessModifiedHybridFuzzyARTMapMode.EXTERNAL ||
param.Mode == ProcessModifiedHybridFuzzyARTMapMode.DUAL)
{
mcoa = new ModifiedClusteringOrderingAlgorithms(param.NoOfClasses);
}
// Analyze max-min of input columns
AnalyzeMaxMin();
// Initialize inputtransformers level 1
inputTransformers1 = new DataTransformChain[Parameter.InputSize];
InitializeInputDataTransform(inputTransformers1, Parameter.InputSize);
// inputtransformation level 2
inputTransformers2 = new DataTransformChain();
inputTransformers2.AppendDataTransform(new ComplementDataTransform(1.0));
// Target column
targetTransform = new DataTransformChain();
targetTransform.AppendDataTransform(new OneOfNDataTransform(Parameter.NoOfClasses));
targetTransform.AppendDataTransform(new ComplementDataTransform(1.0));
// Output transform
outputTransformers = new DataTransformChain();
outputTransformers.AppendDataTransform(new UnComplementDataTransform());
outputTransformers.AppendDataTransform(new ReverseOneOfNDataTransform(Parameter.NoOfClasses));
}
