public static bool Export(Neural network, string pathFile)
{
XmlDocument doc = new XmlDocument();
XmlElement root = doc.CreateElement("Network");
doc.AppendChild(root);
//save number of Input, Hidden, Output Nodes
XmlElement numInput = doc.CreateElement("numInputNodes");
numInput.InnerText = Convert.ToString(network.m_iNumInputNodes);
XmlElement numHidden = doc.CreateElement("numHiddenNodes");
numHidden.InnerText = Convert.ToString(network.m_iNumHiddenNodes);
XmlElement numOutput = doc.CreateElement("numOutputNodes");
numOutput.InnerText = Convert.ToString(network.m_iNumOutputNodes);
root.AppendChild(numInput);
root.AppendChild(numHidden);
root.AppendChild(numOutput);
//save input nodes
XmlElement InputNodes = doc.CreateElement("InputNodes");
for (int i = 0; i <= network.m_iNumInputNodes; i++)
{
XmlElement aInputNode = doc.CreateElement("Input" + Convert.ToString(i + 1));
//save activation func
if (network.m_arInputNodes[i].m_activeFuncType == ActionvationFunction.SIGMOID_FUNCTION)
{
XmlElement actFunc = doc.CreateElement("activateFunc");
actFunc.InnerText = "SIGMOID_FUNCTION";
aInputNode.AppendChild(actFunc);
}
//save weight for in-hid connection
for (int j = 0; j < network.m_iNumHiddenNodes; j++)
{
XmlElement aWeight = doc.CreateElement("InHid" + Convert.ToString(i + 1) + Convert.ToString(j + 1));
aWeight.InnerText = Convert.ToString(network.m_arInputHiddenConn[j, i]);
aInputNode.AppendChild(aWeight);
}
InputNodes.AppendChild(aInputNode);
}
root.AppendChild(InputNodes);
//save hidden nodes
XmlElement HiddenNodes = doc.CreateElement("HiddenNodes");
for (int i = 0; i <= network.m_iNumHiddenNodes; i++)
{
XmlElement aHiddenNode = doc.CreateElement("Hidden" + Convert.ToString(i + 1));
//save activation func
if (network.m_arHiddenNodes[i].m_activeFuncType == ActionvationFunction.SIGMOID_FUNCTION)
{
XmlElement actFunc = doc.CreateElement("activateFunc");
actFunc.InnerText = "SIGMOID_FUNCTION";
aHiddenNode.AppendChild(actFunc);
}
//save weight for hid-out connection
for (int j = 0; j < network.m_iNumOutputNodes; j++)
{
XmlElement aWeight = doc.CreateElement("HidOut" + Convert.ToString(i + 1) + Convert.ToString(j + 1));
aWeight.InnerText = Convert.ToString(network.m_arHiddenOutputConn[j, i]);
aHiddenNode.AppendChild(aWeight);
}
HiddenNodes.AppendChild(aHiddenNode);
}
root.AppendChild(HiddenNodes);
//save output nodes
XmlElement OutputNodes = doc.CreateElement("OutputNodes");
for (int i = 0; i < network.m_iNumOutputNodes; i++)
{
XmlElement aOutputNode = doc.CreateElement("Output" + Convert.ToString(i + 1));
//save activation func
if (network.m_arOutputNodes[i].m_activeFuncType == ActionvationFunction.SIGMOID_FUNCTION)
{
XmlElement actFunc = doc.CreateElement("activateFunc");
actFunc.InnerText = "SIGMOID_FUNCTION";
aOutputNode.AppendChild(actFunc);
}
OutputNodes.AppendChild(aOutputNode);
}
root.AppendChild(OutputNodes);
doc.Save(pathFile);
return true;
}
public static Neural Import(string pathFile)
{
XmlDocument input = new XmlDocument();
Neural loadedNetwork = null;
try
{
input.Load(pathFile);
XmlNode root = input.FirstChild;
//Get number of input, hidden, output nodes
int numInputNodes = Int32.Parse(root.SelectSingleNode("descendant::numInputNodes").InnerText);
int numHiddenNodes = Int32.Parse(root.SelectSingleNode("descendant::numHiddenNodes").InnerText);
int numOutputNodes = Int32.Parse(root.SelectSingleNode("descendant::numOutputNodes").InnerText);
string lags = root.SelectSingleNode("descendant::Lag").InnerText;
//create a network
loadedNetwork = new Neural(numInputNodes, numHiddenNodes, numOutputNodes);
//Get Input Nodes
for (int i = 0; i <= loadedNetwork.m_iNumInputNodes; i++)
{
//get a input node
XmlNode tempNode = root.SelectSingleNode("descendant::Input" + Convert.ToString(i + 1));
//get activation function type
string activationFunc = tempNode.SelectSingleNode("descendant::activateFunc").InnerText;
if (activationFunc.Equals("SIGMOID_FUNCTION"))
{
loadedNetwork.m_arInputNodes[i].m_activeFuncType = ActionvationFunction.SIGMOID_FUNCTION;
}
//get weight
for (int j = 0; j < loadedNetwork.m_iNumHiddenNodes; j++)
{
loadedNetwork.m_arInputHiddenConn[i, j] = Convert.ToDouble(tempNode.SelectSingleNode("descendant::InHid" + Convert.ToString(i + 1) + Convert.ToString(j + 1)).InnerText);
}
}
//Get Hidden Nodes
for (int i = 0; i <= loadedNetwork.m_iNumHiddenNodes; i++)
{
//get a hidden node
XmlNode tempNode = root.SelectSingleNode("descendant::Hidden" + Convert.ToString(i + 1));
//get activation function type
string activationFunc = tempNode.SelectSingleNode("descendant::activateFunc").InnerText;
if (activationFunc.Equals("SIGMOID_FUNCTION"))
{
loadedNetwork.m_arHiddenNodes[i].m_activeFuncType = ActionvationFunction.SIGMOID_FUNCTION;
}
}
//Get Output Nodes
for (int i = 0; i < loadedNetwork.m_iNumOutputNodes; i++)
{
//get a output node
XmlNode tempNode = root.SelectSingleNode("descendant::Output" + Convert.ToString(i + 1));
//get activation function type
string activationFunc = tempNode.SelectSingleNode("descendant::activateFunc").InnerText;
if (activationFunc.Equals("SIGMOID_FUNCTION"))
{
loadedNetwork.m_arOutputNodes[i].m_activeFuncType = ActionvationFunction.SIGMOID_FUNCTION;
}
}
}
catch (Exception e)
{
Console.WriteLine(e.Message);
return null;
}
return loadedNetwork;
}
public static bool Export(Neural network, string pathFile)
{
XmlDocument doc = new XmlDocument();
XmlElement root = doc.CreateElement("Network");
doc.AppendChild(root);
//save number of Input, Hidden, Output Nodes
XmlElement numInput = doc.CreateElement("numInputNodes");
numInput.InnerText = Convert.ToString(network.m_iNumInputNodes);
XmlElement numHidden = doc.CreateElement("numHiddenNodes");
numHidden.InnerText = Convert.ToString(network.m_iNumHiddenNodes);
XmlElement numOutput = doc.CreateElement("numOutputNodes");
numOutput.InnerText = Convert.ToString(network.m_iNumOutputNodes);
root.AppendChild(numInput);
root.AppendChild(numHidden);
root.AppendChild(numOutput);
//save input nodes
XmlElement InputNodes = doc.CreateElement("InputNodes");
for (int i = 0; i <= network.m_iNumInputNodes; i++)
{
XmlElement aInputNode = doc.CreateElement("Input" + Convert.ToString(i + 1));
//save activation func
if (network.m_arInputNodes[i].m_activeFuncType == ActionvationFunction.SIGMOID_FUNCTION)
{
XmlElement actFunc = doc.CreateElement("activateFunc");
actFunc.InnerText = "SIGMOID_FUNCTION";
aInputNode.AppendChild(actFunc);
}
//save weight for in-hid connection
for (int j = 0; j < network.m_iNumHiddenNodes; j++)
{
XmlElement aWeight = doc.CreateElement("InHid" + Convert.ToString(i + 1) + Convert.ToString(j + 1));
aWeight.InnerText = Convert.ToString(network.m_arInputHiddenConn[j, i]);
aInputNode.AppendChild(aWeight);
}
InputNodes.AppendChild(aInputNode);
}
root.AppendChild(InputNodes);
//save hidden nodes
XmlElement HiddenNodes = doc.CreateElement("HiddenNodes");
for (int i = 0; i <= network.m_iNumHiddenNodes; i++)
{
XmlElement aHiddenNode = doc.CreateElement("Hidden" + Convert.ToString(i + 1));
//save activation func
if (network.m_arHiddenNodes[i].m_activeFuncType == ActionvationFunction.SIGMOID_FUNCTION)
{
XmlElement actFunc = doc.CreateElement("activateFunc");
actFunc.InnerText = "SIGMOID_FUNCTION";
aHiddenNode.AppendChild(actFunc);
}
//save weight for hid-out connection
for (int j = 0; j < network.m_iNumOutputNodes; j++)
{
XmlElement aWeight = doc.CreateElement("HidOut" + Convert.ToString(i + 1) + Convert.ToString(j + 1));
aWeight.InnerText = Convert.ToString(network.m_arHiddenOutputConn[j, i]);
aHiddenNode.AppendChild(aWeight);
}
HiddenNodes.AppendChild(aHiddenNode);
}
root.AppendChild(HiddenNodes);
//save output nodes
XmlElement OutputNodes = doc.CreateElement("OutputNodes");
for (int i = 0; i < network.m_iNumOutputNodes; i++)
{
XmlElement aOutputNode = doc.CreateElement("Output" + Convert.ToString(i + 1));
//save activation func
if (network.m_arOutputNodes[i].m_activeFuncType == ActionvationFunction.SIGMOID_FUNCTION)
{
XmlElement actFunc = doc.CreateElement("activateFunc");
actFunc.InnerText = "SIGMOID_FUNCTION";
aOutputNode.AppendChild(actFunc);
}
OutputNodes.AppendChild(aOutputNode);
}
root.AppendChild(OutputNodes);
doc.Save(pathFile);
return(true);
}
public static Neural Import(string pathFile)
{
XmlDocument input = new XmlDocument();
Neural loadedNetwork = null;
try
{
input.Load(pathFile);
XmlNode root = input.FirstChild;
//Get number of input, hidden, output nodes
int numInputNodes = Int32.Parse(root.SelectSingleNode("descendant::numInputNodes").InnerText);
int numHiddenNodes = Int32.Parse(root.SelectSingleNode("descendant::numHiddenNodes").InnerText);
int numOutputNodes = Int32.Parse(root.SelectSingleNode("descendant::numOutputNodes").InnerText);
string lags = root.SelectSingleNode("descendant::Lag").InnerText;
//create a network
loadedNetwork = new Neural(numInputNodes, numHiddenNodes, numOutputNodes);
//Get Input Nodes
for (int i = 0; i <= loadedNetwork.m_iNumInputNodes; i++)
{
//get a input node
XmlNode tempNode = root.SelectSingleNode("descendant::Input" + Convert.ToString(i + 1));
//get activation function type
string activationFunc = tempNode.SelectSingleNode("descendant::activateFunc").InnerText;
if (activationFunc.Equals("SIGMOID_FUNCTION"))
{
loadedNetwork.m_arInputNodes[i].m_activeFuncType = ActionvationFunction.SIGMOID_FUNCTION;
}
//get weight
for (int j = 0; j < loadedNetwork.m_iNumHiddenNodes; j++)
{
loadedNetwork.m_arInputHiddenConn[i, j] = Convert.ToDouble(tempNode.SelectSingleNode("descendant::InHid" + Convert.ToString(i + 1) + Convert.ToString(j + 1)).InnerText);
}
}
//Get Hidden Nodes
for (int i = 0; i <= loadedNetwork.m_iNumHiddenNodes; i++)
{
//get a hidden node
XmlNode tempNode = root.SelectSingleNode("descendant::Hidden" + Convert.ToString(i + 1));
//get activation function type
string activationFunc = tempNode.SelectSingleNode("descendant::activateFunc").InnerText;
if (activationFunc.Equals("SIGMOID_FUNCTION"))
{
loadedNetwork.m_arHiddenNodes[i].m_activeFuncType = ActionvationFunction.SIGMOID_FUNCTION;
}
}
//Get Output Nodes
for (int i = 0; i < loadedNetwork.m_iNumOutputNodes; i++)
{
//get a output node
XmlNode tempNode = root.SelectSingleNode("descendant::Output" + Convert.ToString(i + 1));
//get activation function type
string activationFunc = tempNode.SelectSingleNode("descendant::activateFunc").InnerText;
if (activationFunc.Equals("SIGMOID_FUNCTION"))
{
loadedNetwork.m_arOutputNodes[i].m_activeFuncType = ActionvationFunction.SIGMOID_FUNCTION;
}
}
}
catch (Exception e)
{
Console.WriteLine(e.Message);
return(null);
}
return(loadedNetwork);
}
static void Main(string[] args)
{
Neural neural = new Neural(1, 2, 1);
double learningRate = 0.7;
double moment = 0.4;
double maxEpouch = 10000;
double expectError = 0.00001;
string fileName = @"E:\PROJECT\FINAL PROJECT\Other\Test\fuel.txt";
List<double> sample = new List<double>();
System.IO.StreamReader file = null;
string line = null;
int counter = 0;
bool isFormatFileRight = true;
int beginRow = 1;
int endRow = 71;
int columnSelected = 1;
int idxRow = 0;
try
{
file = new System.IO.StreamReader(fileName);
while ((line = file.ReadLine()) != null)
{
idxRow++;
if (idxRow < beginRow || idxRow > endRow)
continue;
char[] delimiterChars = { ' ', ',' };
string[] words = line.Split(delimiterChars);
if (columnSelected <= words.Length)
{
sample.Add(Double.Parse(words[columnSelected - 1]));
}
else
{
isFormatFileRight = false;
break;
}
}
}
catch (System.OutOfMemoryException outOfMemory)
{
sample = null;
}
double max = sample.Max();
double min = sample.Min();
int count = sample.Count;
double[] series = new double[count];
List<double> sample2 = new List<double>();
for (int i = 0; i < count; i++)
{
double a = sample.ElementAt(i);
double b = (a - min) / (max - min) * (0.99 - 0.01) + 0.01;
series[i] = b;
sample2.Add(b);
}
NeuralTraining training = new NeuralTraining();
training.s_Network = neural;
//training.Rprop_Run(sample2, null);
training.Bp_Run(sample2, null, 0.7, 0.4);
int x = 0;
}
static void Main(string[] args)
{
Neural neural = new Neural(1, 2, 1);
double learningRate = 0.7;
double moment = 0.4;
double maxEpouch = 10000;
double expectError = 0.00001;
string fileName = @"E:\PROJECT\FINAL PROJECT\Other\Test\fuel.txt";
List <double> sample = new List <double>();
System.IO.StreamReader file = null;
string line = null;
int counter = 0;
bool isFormatFileRight = true;
int beginRow = 1;
int endRow = 71;
int columnSelected = 1;
int idxRow = 0;
try
{
file = new System.IO.StreamReader(fileName);
while ((line = file.ReadLine()) != null)
{
idxRow++;
if (idxRow < beginRow || idxRow > endRow)
{
continue;
}
char[] delimiterChars = { ' ', ',' };
string[] words = line.Split(delimiterChars);
if (columnSelected <= words.Length)
{
sample.Add(Double.Parse(words[columnSelected - 1]));
}
else
{
isFormatFileRight = false;
break;
}
}
}
catch (System.OutOfMemoryException outOfMemory)
{
sample = null;
}
double max = sample.Max();
double min = sample.Min();
int count = sample.Count;
double[] series = new double[count];
List <double> sample2 = new List <double>();
for (int i = 0; i < count; i++)
{
double a = sample.ElementAt(i);
double b = (a - min) / (max - min) * (0.99 - 0.01) + 0.01;
series[i] = b;
sample2.Add(b);
}
NeuralTraining training = new NeuralTraining();
training.s_Network = neural;
//training.Rprop_Run(sample2, null);
training.Bp_Run(sample2, null, 0.7, 0.4);
int x = 0;
}