public void TestXMLSerializationAndDeserialization()
{
IActivationFunction func = new BipolarSigmoid(2.0);
var layerCounts = new List <int>();
layerCounts.Add(5);
layerCounts.Add(10);
layerCounts.Add(1);
var network = new Network(5, 3, layerCounts, func, new RandomInitializer());
var serializer = new DataContractSerializer(typeof(Network), new[] { func.GetType(), typeof(RandomInitializer) });
var settings = new XmlWriterSettings();
settings.Indent = true;
settings.IndentChars = " ";//4 spaces
string file = "serialized_net.xml";
using (var writer = XmlWriter.Create(file, settings))
{
serializer.WriteObject(writer, network);
}
Network newNet = (Network)serializer.ReadObject(XmlReader.Create(file));
Assert.IsNotNull(newNet);
Assert.AreEqual(5, newNet.Inputs);
Assert.AreEqual(3, newNet.LayersCount);
Assert.IsTrue(0 < newNet[0][0].Weights[0]);
Assert.AreEqual(typeof(BipolarSigmoid), newNet[0][0].ActivationFunction.GetType());
}
public void TestXMLSerialization()
{
IActivationFunction func = new BipolarSigmoid(2.0);
var layerCounts = new List <int>();
layerCounts.Add(5);
layerCounts.Add(10);
layerCounts.Add(1);
var network = new Network(5, 3, layerCounts, func, new RandomInitializer());
var serializer = new DataContractSerializer(typeof(Network), new[] { func.GetType(), typeof(RandomInitializer) });
var fileName = "serialized_net.xml";
var settings = new XmlWriterSettings();
settings.Indent = true;
settings.IndentChars = " ";//4 spaces
using (var writer = XmlWriter.Create(fileName, settings))
{
serializer.WriteObject(writer, network);
}
var fInfo = new FileInfo(fileName);
Assert.IsTrue(fInfo.Exists);
Assert.IsTrue(fInfo.Length > 0);
}
public void Load(byte[] data)
{
var memStream = new MemoryStream();
var binForm = new BinaryFormatter();
memStream.Write(data, 0, data.Length);
memStream.Seek(0, SeekOrigin.Begin);
var obj = (SaveNetworkModel)binForm.Deserialize(memStream);
var layers = new List <ILayer>();
Layers = null;
foreach (var layer in obj.Layers)
{
var function = new BipolarSigmoid();
var relu = new ELU();
if (layer.Type == LayerType.Convolution)
{
layers.Add(new ConvolutionalLayer(layer.ActivationType, layer.OutputLength, layer.KernelSize));
}
if (layer.Type == LayerType.MaxPoolingLayer)
{
layers.Add(new MaxPoolingLayer(layer.KernelSize));
}
if (layer.Type == LayerType.FullyConnected)
{
layers.Add(new FullyConnectedLayer(layer.OutputLength, layer.ActivationType));
}
}
InitLayers(obj.InputWidth, obj.InputHeight, layers.ToArray());
for (var l = 0; l < Layers.Length; l++)
{
if (Layers[l].Type == LayerType.Convolution)
{
var layer = (ConvolutionalLayer)Layers[l];
for (var n = 0; n < layer.NeuronsCount; n++)
{
layer.Neurons[n].Weights = obj.Layers[l].ConvNeurons[n].Weights;
}
}
if (Layers[l].Type == LayerType.FullyConnected)
{
var layer = (FullyConnectedLayer)Layers[l];
for (var n = 0; n < layer.NeuronsCount; n++)
{
layer.Neurons[n].Weights = obj.Layers[l].FullyConnectedNeurons[n].Weights;
}
}
}
}
public void BipolarSigmoidPrimeTest()
{
var a = new Matrix(2, 2);
a.InRandomize();
var b = a.Duplicate();
a = new BipolarSigmoid().Backward(a);
b.InMap((x) => 0.5 * (1 + (-1 + 2 / (1 + Math.Exp(-x)))) * (1 - (-1 + 2 / (1 + Math.Exp(-x)))));
Assert.IsTrue(Math.Abs(a.FrobeniusNorm() - b.FrobeniusNorm()) < 0.1,
new BipolarSigmoid().Type().ToString() + " Derivative.");
}
public void BipolarSigmoidTest()
{
var a = new Matrix(2, 2);
a.InRandomize();
var b = a.Duplicate();
a = new BipolarSigmoid().Forward(a);
b.InMap((x) => - 1 + 2 / (1 + Math.Exp(-x)));
Assert.IsTrue(Math.Abs(a.FrobeniusNorm() - b.FrobeniusNorm()) < 0.1,
new BipolarSigmoid().Type().ToString() + " Activation.");
}
public void TestNetworkWorkingAtAll_TryAndOperation()
{
IActivationFunction func = new BipolarSigmoid(2.0);
var layerCounts = new List <int>();
layerCounts.Add(2);
layerCounts.Add(2);
layerCounts.Add(1);
var network = new Network(2, 3, layerCounts, func, new RandomInitializer());
//double[] input = new double[] { -3, -2, -1, 0, 1, 2, 3 };
//double[] output = new double[] { 9, 4, 1, 0, 1, 4, 9 };
double[][] input = new double[][] { new[] { 0.0, 0.0 }, new[] { 1.0, 0.0 }, new[] { 0.0, 1.0 }, new[] { 1.0, 1.0 } };
double[] output = new double[] { 0.0, 1.0, 1.0, 1.0 };
//double[] input = new double[] { 1,2,3,4,5,6,7,8,9 };
//double[] output = new double[]{ 0,4,7,9,10,9,7,4,0 };
var learner = new Teacher(network);
learner.Rate = 0.2;
learner.Momentum = 0.0;
List <double[]> netInputs = new List <double[]>();
List <double[]> netOutputs = new List <double[]>();
for (int i = 0; i < input.Length; i++)
{
netInputs.Add(input[i]);
netOutputs.Add(new double[] { output[i] });
}
for (int i = 0; i < 1000; i++)
{
learner.TeachOnSamples(netInputs, netOutputs);
}
double[] aproximated = new double[input.Length];
Console.WriteLine("Try those Values in Excel or whatever");
using (var writer = new StreamWriter("data_dump.txt"))
{
for (int i = 0; i < aproximated.Length; i++)
{
aproximated[i] = network.ComputeOutputVector(input[i])[0];
writer.WriteLine((aproximated[i]).ToString());
}
}
}
public Layer(int nCount, int index, ActivationSettings activationSettings)
{
NCount = nCount;
Index = index;
ActivationType = activationSettings.Type();
// Activation Setup
switch (activationSettings.Type())
{
case EActivationType.Invalid:
Activation = null;
throw new ArgumentException("Activation Type Invalid.");
case EActivationType.Arctan:
Activation = new Arctan();
break;
case EActivationType.BinaryStep:
Activation = new BinaryStep();
break;
case EActivationType.BipolarSigmoid:
Activation = new BipolarSigmoid();
break;
case EActivationType.ELU:
Activation = new ELU((ELUSettings)activationSettings);
break;
case EActivationType.HardSigmoid:
Activation = new HardSigmoid();
break;
case EActivationType.HardTanh:
Activation = new HardTanh();
break;
case EActivationType.Identity:
Activation = new Identity();
break;
case EActivationType.Logit:
Activation = new Logit();
break;
case EActivationType.LReLU:
Activation = new LReLU((LReLUSettings)activationSettings);
break;
case EActivationType.Mish:
Activation = new Mish();
break;
case EActivationType.ReLU:
Activation = new ReLU();
break;
case EActivationType.SeLU:
Activation = new SeLU();
break;
case EActivationType.Sigmoid:
Activation = new Sigmoid();
break;
case EActivationType.Softmax:
Activation = new Softmax();
break;
case EActivationType.Softplus:
Activation = new Softplus();
break;
case EActivationType.Softsign:
Activation = new Softsign();
break;
case EActivationType.Tanh:
Activation = new Tanh();
break;
default:
throw new ArgumentException("Activation Type Invalid.");
}
}
private void AssignValue()
{
switch (initFuncComboBox.SelectedIndex)
{
case (0):
Init = new RandomInitializer();
break;
case (1):
Init = new OptimalRangeRandomInitializer(Activ = new BipolarSigmoid((double)alphaNumeric.Value));
break;
case (2):
Init = new ConstInitializer(double.Parse(constValueTextBox.Text));
break;
default:
Init = null;
break;
}
if (activFuncComboBox.SelectedIndex == 1)
{
Activ = new BipolarSigmoid((double)alphaNumeric.Value);
}
else
{
Activ = new Sigmoid((double)alphaNumeric.Value);
}
LearnerRate = (double)rateNumeric.Value;
LearnerMomentum = (double)momentumNumeric.Value;
int firstLayer = (int)windowSize.Value;
firstLayer += AddIndicators();
Layer = new List <int> {
firstLayer
};
if (layerCountCheckBox.Checked)
{
for (int i = 0; i < (int)layersNumeric.Value - 2; i++)
{
Layer.Add((int)windowSize.Value * 2);
}
}
else
{
for (int i = 0; i < (int)layersNumeric.Value - 2; i++)
{
var neuron = new NeuronCounts {
Text = "Layer no. " + i + " count"
};
DialogResult res = neuron.ShowDialog(this);
if (res == DialogResult.OK)
{
Layer.Add(neuron.Value);
}
}
}
Layer.Add(1);
int iterations;
if (!int.TryParse(iterationsTextBox.Text, out iterations))
{
iterations = 1000;
}
IterationsCount = iterations;
}
private void tests()
{
//using (var writer = new StreamWriter("1hidelayer.txt"))
//{
// Init = new RandomInitializer();
// Activ = new BipolarSigmoid((double)2.0);
// LearnerRate = 0.3;
// LearnerMomentum = 0.0;
// string column = "window|hide layer 1|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1";
// writer.WriteLine(column);
// for (int j = 2; j < 10; j++)
// {
// for (int i = 1; i < 21; i++)
// {
// string toFile = j + "|" + i + "|";
// Layer = new List<int> { j };
// Layer.Add(i);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// }
// }
//}
using (var writer = new StreamWriter("2hidelayer100.txt"))
{
Init = new OptimalRangeRandomInitializer(Activ = new BipolarSigmoid(2.0));
LearnerRate = 0.3;
LearnerMomentum = 0.0;
IterationsCount = 100;
string column = "window|hide layer 1|hide layer 2|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1";
writer.WriteLine(column);
for (int j = 2; j < 10; j++)
{
for (int i = 1; i < 21; i++)
{
for (int k = 1; k < 21; k++)
{
string toFile = j + "|" + i + "|" + k + "|";
Layer = new List <int> {
j
};
Layer.Add(i);
Layer.Add(k);
Layer.Add(1);
var a = this.Owner as MainForm;
if (a != null)
{
a.SetSettings();
toFile += a.Teach(false).ToString("F6") + "|";
toFile += a.Predict(false, true).ToString("F6") + "|";
toFile += a.Predict(false, true, 10).ToString("F6") + "|";
toFile += a.Predict(false, true, 5).ToString("F6") + "|";
toFile += a.Predict(false, true, 1).ToString("F6") + "|";
toFile += a.Predict(false, false).ToString("F6") + "|";
toFile += a.Predict(false, false, 10).ToString("F6") + "|";
toFile += a.Predict(false, false, 5).ToString("F6") + "|";
toFile += a.Predict(false, false, 1).ToString("F6");
}
writer.WriteLine(toFile);
}
}
}
}
using (var writer = new StreamWriter("2hidelayer200.txt"))
{
Init = new OptimalRangeRandomInitializer(Activ = new BipolarSigmoid(2.0));
LearnerRate = 0.3;
LearnerMomentum = 0.0;
IterationsCount = 500;
string column = "window|hide layer 1|hide layer 2|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1";
writer.WriteLine(column);
for (int j = 2; j < 10; j++)
{
for (int i = 1; i < 21; i++)
{
for (int k = 1; k < 21; k++)
{
string toFile = j + "|" + i + "|" + k + "|";
Layer = new List <int> {
j
};
Layer.Add(i);
Layer.Add(k);
Layer.Add(1);
var a = this.Owner as MainForm;
if (a != null)
{
a.SetSettings();
toFile += a.Teach(false).ToString("F6") + "|";
toFile += a.Predict(false, true).ToString("F6") + "|";
toFile += a.Predict(false, true, 10).ToString("F6") + "|";
toFile += a.Predict(false, true, 5).ToString("F6") + "|";
toFile += a.Predict(false, true, 1).ToString("F6") + "|";
toFile += a.Predict(false, false).ToString("F6") + "|";
toFile += a.Predict(false, false, 10).ToString("F6") + "|";
toFile += a.Predict(false, false, 5).ToString("F6") + "|";
toFile += a.Predict(false, false, 1).ToString("F6");
}
writer.WriteLine(toFile);
}
}
}
}
//using (var writer = new StreamWriter("1hidelayerRandomInit.txt"))
//{
// Init = new RandomInitializer();
// Activ = new BipolarSigmoid((double)2.0);
// LearnerRate = 0.3;
// LearnerMomentum = 0.0;
// IterationsCount = 500;
// writer.WriteLine("func activ -BipolarSigmoid((double)2.0) , init - RandomInitializer() 500 iteracji");
// string column = "window|hide layer 1|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1";
// writer.WriteLine(column);
// for (int j = 2; j < 10; j++)
// {
// for (int i = 1; i < 21; i++)
// {
// string toFile = j + "|" + i + "|";
// Layer = new List<int> { j };
// Layer.Add(i);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// }
// }
//}
//using (var writer = new StreamWriter("1hidelayerConstInit.txt"))
//{
// Init = new ConstInitializer(5);
// Activ = new BipolarSigmoid((double)2.0);
// LearnerRate = 0.3;
// LearnerMomentum = 0.0;
// IterationsCount = 500;
// string column = "window|hide layer 1|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1";
// writer.WriteLine("func activ -BipolarSigmoid((double)2.0) , init - ConstInitializer(5) 500 iteracji");
// writer.WriteLine(column);
// for (int j = 2; j < 10; j++)
// {
// for (int i = 1; i < 21; i++)
// {
// string toFile = j + "|" + i + "|";
// Layer = new List<int> { j };
// Layer.Add(i);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// }
// }
//}
//using (var writer = new StreamWriter("1hidelayerSigmoid.txt"))
//{
// Init = new RandomInitializer();
// Activ = new Sigmoid(2.0);
// LearnerRate = 0.3;
// LearnerMomentum = 0.0;
// IterationsCount = 500;
// string column = "window|hide layer 1|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1";
// writer.WriteLine("func activ -Sigmoid(2.0) , init - RandomInitializer() 500 iteracji");
// writer.WriteLine(column);
// for (int j = 2; j < 10; j++)
// {
// for (int i = 1; i < 21; i++)
// {
// string toFile = j + "|" + i + "|";
// Layer = new List<int> { j };
// Layer.Add(i);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// }
// }
//}
// using (var writer = new StreamWriter("wskazniki.txt"))
// {
// Init = new OptimalRangeRandomInitializer(Activ = new BipolarSigmoid(2.0));
// ////Activ = new BipolarSigmoid((double)2.0);
// LearnerRate = 0.3;
// LearnerMomentum = 0.0;
// IterationsCount = 100;
// writer.WriteLine("indicators|argument|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1");
// for (int j = 0; j < 6; j++)
// {
// for (int i = 0; i < 30; i++)
// {
// string toFile = "";
// switch (j)
// {
// case 0:
// Indicator.Clear();
// Indicator.Add(new KeyValuePair<Indi, int>(Indicators.Indicators.SMA, i));
// toFile += "sma|";
// break;
// case 1:
// Indicator.Clear();
// Indicator.Add(new KeyValuePair<Indi, int>(Indicators.Indicators.WMA, i));
// toFile += "wma|";
// break;
// case 2:
// Indicator.Clear();
// Indicator.Add(new KeyValuePair<Indi, int>(Indicators.Indicators.EMA, i));
// toFile += "ema|";
// break;
// case 3:
// Indicator.Clear();
// Indicator.Add(new KeyValuePair<Indi, int>(Indicators.Indicators.ROC, i));
// toFile += "roc|";
// break;
// case 4:
// Indicator.Clear();
// Indicator.Add(new KeyValuePair<Indi, int>(Indicators.Indicators.MACD, i));
// toFile += "macd|";
// break;
// case 5:
// Indicator.Clear();
// Indicator.Add(new KeyValuePair<Indi, int>(Indicators.Indicators.Oscillator, 3 + i));
// toFile += "oscil|";
// break;
// }
// toFile += i + "|";
// Layer = new List<int> { 4 };
// Layer.Add(12);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// }
// }
// }
// using (var writer = new StreamWriter("parametry.txt"))
// {
// writer.WriteLine("rate|momentum|alpha|teach error|predic error 30%|10 values|5 |1|predic error 30%|10 values|5 |1");
// double alpha = 0;
// LearnerRate = 0.0;
// LearnerMomentum = 0.0;
// IterationsCount = 100;
// for (int j = 0; j < 50; j++)
// {
// Init = new OptimalRangeRandomInitializer(Activ = new BipolarSigmoid(2.0));
// //Activ = new BipolarSigmoid((double)2.0);
// string toFile = LearnerRate + "|" + LearnerMomentum + "|2.0|";
// Layer = new List<int> { 3 };
// Layer.Add(12);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// LearnerRate += 0.05;
// }
// for (int j = 0; j < 50; j++)
// {
// Init = new OptimalRangeRandomInitializer(Activ = new BipolarSigmoid(2.0));
// //Activ = new BipolarSigmoid((double)2.0);
// LearnerRate = 0.3;
// string toFile = LearnerRate + "|" + LearnerMomentum + "|2.0|";
// Layer = new List<int> { 3 };
// Layer.Add(12);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// LearnerMomentum += 0.05;
// }
// for (int j = 0; j < 50; j++)
// {
// Init = new OptimalRangeRandomInitializer(Activ = new BipolarSigmoid(alpha));
// //Activ = new BipolarSigmoid((double)2.0);
// LearnerRate = 0.3;
// LearnerMomentum = 0.0;
// string toFile = LearnerRate + "|" + LearnerMomentum + "|" + alpha + "|";
// Layer = new List<int> { 3 };
// Layer.Add(12);
// Layer.Add(1);
// var a = this.Owner as MainForm;
// if (a != null)
// {
// a.SetSettings();
// toFile += a.Teach(false).ToString("F6") + "|";
// toFile += a.Predict(false, true).ToString("F6") + "|";
// toFile += a.Predict(false, true, 10).ToString("F6") + "|";
// toFile += a.Predict(false, true, 5).ToString("F6") + "|";
// toFile += a.Predict(false, true, 1).ToString("F6") + "|";
// toFile += a.Predict(false, false).ToString("F6") + "|";
// toFile += a.Predict(false, false, 10).ToString("F6") + "|";
// toFile += a.Predict(false, false, 5).ToString("F6") + "|";
// toFile += a.Predict(false, false, 1).ToString("F6");
// }
// writer.WriteLine(toFile);
// alpha += 0.1;
// }
// }
}
public void TestTimeSeriesPrediction()
{
const int windowSize = 5;
const int predictionSize = 1;
IActivationFunction func = new BipolarSigmoid(2.0);
var layerCounts = new List <int>();
layerCounts.Add(windowSize);
layerCounts.Add(2 * windowSize);
layerCounts.Add(1);
var network = new Network(5, 3, layerCounts, func, new RandomInitializer());
//double[] input = new double[] { -3, -2, -1, 0, 1, 2, 3 };
//double[] output = new double[] { 9, 4, 1, 0, 1, 4, 9 };
double[] readedData = new double[50];
double[] data;
double min, max;
double range;
using (var reader = new StreamReader(File.OpenRead("sinusoid.csv")))
{
int i = 0;
readedData[i] = double.Parse(reader.ReadLine(), System.Globalization.CultureInfo.InvariantCulture);
min = readedData[i];
max = readedData[i];
try
{
for (i = 1; i < 50; i++)
{
readedData[i] = double.Parse(reader.ReadLine(), System.Globalization.CultureInfo.InvariantCulture);
if (min > readedData[i])
{
min = readedData[i];
}
if (max < readedData[i])
{
max = readedData[i];
}
}
}
catch (Exception)
{
}
if (i > 0)
{
data = new double[i];
Array.Copy(readedData, data, i);
}
else
{
return;
}
}
range = (max - min);
int learningSamples = data.Length - windowSize - predictionSize;
double[][] input = new double[learningSamples][];
double[][] output = new double[learningSamples][];
for (int i = 0; i < learningSamples; i++)
{
input[i] = new double[windowSize];
output[i] = new double[1];
for (int j = 0; j < windowSize; j++)
{
input[i][j] = TransformData(data[i + j], min, range);
}
output[i][0] = TransformData(data[i + windowSize], min, range);
}
//double[] input = new double[] { 1,2,3,4,5,6,7,8,9 };
//double[] output = new double[]{ 0,4,7,9,10,9,7,4,0 };
var learner = new Teacher(network);
learner.Rate = 0.1;
learner.Momentum = 0.0;
var solution = new double[data.Length - windowSize, 2];
var netInput = new double[windowSize];
for (int i = 0; i < 1000; i++)
{
var ins = new List <double[]>();
var outs = new List <double[]>();
for (int ii = 0; ii < input.Length; ii++)
{
ins.Add(input[ii]);
outs.Add(output[ii]);
}
learner.TeachOnSamples(ins, outs);
for (int j = 0; j < data.Length - windowSize; j++)
{
for (int k = 0; k < windowSize; k++)
{
netInput[k] = TransformData(data[j + k], min, range);
}
solution[j, 1] = TransformBack(network.ComputeOutputVector(netInput)[0], min, max);//(network.ComputeOutputVector(netInput)[0]) / range + min;
}
}
double[] aproximated = new double[data.Length - windowSize];
Console.WriteLine("Try those Values in Excel or whatever");
using (var writer = new StreamWriter("data_dump_time_series.txt"))
{
for (int i = 0; i < aproximated.Length; i++)
{
aproximated[i] = solution[i, 1];
writer.WriteLine((aproximated[i]).ToString());
}
}
}
