public BackpropagationNetwork network(int trainInVectorDimension, int trainOutVectorDimension)
{
this.hiddenLayerList = HiddenLayerList();
ActivationLayer inputLayer = new LinearLayer(trainInVectorDimension);
ActivationLayer outputLayer = new SigmoidLayer(trainOutVectorDimension);
BackpropagationConnector bpc0 = new BackpropagationConnector(inputLayer, this.hiddenLayerList[0]);
for (int i = 1; i < this.hiddenLayerList.Count; i++)
{
bpc0 = new BackpropagationConnector(this.hiddenLayerList[i - 1], this.hiddenLayerList[i]);
}
bpc0 = new BackpropagationConnector(this.hiddenLayerList[this.hiddenLayerList.Count - 1], outputLayer);
BackpropagationNetwork network = new BackpropagationNetwork(inputLayer, outputLayer);
/*ActivationLayer inputLayer = hiddenLayerList[0];
* ActivationLayer outputLayer = hiddenLayerList[hiddenLayerList.Count - 1];
*
* if(hiddenLayerList.Count != 2)
* {
* BackpropagationConnector bpc0 = new BackpropagationConnector(inputLayer, this.hiddenLayerList[1]);
* for (int i = 2; i < this.hiddenLayerList.Count - 1; i++)
* {
* bpc0 = new BackpropagationConnector(this.hiddenLayerList[i - 1], this.hiddenLayerList[i]);
* }
* bpc0 = new BackpropagationConnector(this.hiddenLayerList[this.hiddenLayerList.Count - 2], outputLayer);
* }
*
* BackpropagationNetwork network = new BackpropagationNetwork(inputLayer, outputLayer);*/
network.SetLearningRate(this.learningRate);
return(network);
}
public void TestMethod1()
{
// 创建输入层、隐层和输出层
var inputLayer = new LinearLayer(1);
var hiddenLayer = new LinearLayer(5);
var outputLayer = new LinearLayer(1);
// 创建层之间的关联
new BackpropagationConnector(inputLayer, hiddenLayer, ConnectionMode.Complete);
new BackpropagationConnector(hiddenLayer, outputLayer, ConnectionMode.Complete);
// 创建神经网络
var network = new BackpropagationNetwork(inputLayer, outputLayer);
//network.SetLearningRate(new LinearFunction(0.1, 0.6));
network.Initialize();
// 训练
var ran = new Random();
for (var i = 0; i < 100; i++)
{
var inputVector = new double[] { i };
var outputVector = new double[] { Math.PI *i };
var trainingSample = new TrainingSample(inputVector, outputVector);
network.Learn(trainingSample, i, 100);
}
// 预测
var testInput = new double[] { 1 };
var testOutput = network.Run(testInput);
Console.WriteLine(testOutput[0]);
}
public void LabTest1()
{
var inputLayer = new LinearLayer(5);
var hiddenLayer = new TanhLayer(neuronCount);
var outputLayer = new TanhLayer(2);
new BackpropagationConnector(inputLayer, hiddenLayer);
new BackpropagationConnector(hiddenLayer, outputLayer);
_xorNetwork = new BackpropagationNetwork(inputLayer, outputLayer);
_xorNetwork.SetLearningRate(learningRate);
var trainingSet = new TrainingSet(5, 2);
trainingSet.Add(new TrainingSample(new double[] { 0, 0, 0, 0, 0 }, new double[] { 0, 0 }));
trainingSet.Add(new TrainingSample(new double[] { 0, 0, 0, 1, 0 }, new double[] { 3, 3 }));
trainingSet.Add(new TrainingSample(new double[] { 0, 0, 1, 0, 0 }, new double[] { 2, 2 }));
trainingSet.Add(new TrainingSample(new double[] { 0, 0, 1, 1, 0 }, new double[] { 2, 3 }));
trainingSet.Add(new TrainingSample(new double[] { 0, 1, 0, 0, 0 }, new double[] { 1, 1 }));
trainingSet.Add(new TrainingSample(new double[] { 0, 1, 0, 1, 0 }, new double[] { 1, 3 }));
trainingSet.Add(new TrainingSample(new double[] { 0, 1, 1, 0, 0 }, new double[] { 1, 2 }));
trainingSet.Add(new TrainingSample(new double[] { 0, 1, 1, 1, 0 }, new double[] { 1, 3 }));
trainingSet.Add(new TrainingSample(new double[] { 22, 1, 1, 1, 22 }, new double[] { 1, 3 }));
_errorList = new double[cycles];
//_xorNetwork.EndEpochEvent += EndEpochEvent;
_xorNetwork.Learn(trainingSet, cycles);
var result = _xorNetwork.Run(new double[] { 0, 0, 1, 1, 0 });
}
/// <summary>
/// Create a new adam network trainer
/// </summary>
/// <param name="TargetNetwork">Network to train</param>
/// <param name="TrainingDataset">Data sets used for training</param>
/// <param name="LearningRate">Initial learning rate, default = 0.1</param>
public AdamTrainer(BackpropagationNetwork TargetNetwork, IOMetaDataSet <double[]> TrainingDataset, double LearningRate = 0.1)
{
this.TargetNetwork = TargetNetwork;
this.TrainingDataset = TrainingDataset;
adamDecent = new AdamDecent(LearningRate);
}
public override void Train(IForecastingDataSets datasets)
{
OnStartRunning(new ComponentRunEventArgs(datasets));
AnnModelParameter para = mParameter as AnnModelParameter;
LinearLayer inputLayer = new LinearLayer(datasets.InputData[0].Length);
SigmoidLayer hiddenLayer = new SigmoidLayer(para.HiddenNeuronsCount[0]);
SigmoidLayer outputLayer = new SigmoidLayer(1);
new BackpropagationConnector(inputLayer, hiddenLayer).Initializer = new RandomFunction(0d, 0.3d);
new BackpropagationConnector(hiddenLayer, outputLayer).Initializer = new RandomFunction(0d, 0.3d);
network = new BackpropagationNetwork(inputLayer, outputLayer);
network.SetLearningRate(para.LearningRate);
network.JitterEpoch = para.JitterEpoch;
network.JitterNoiseLimit = para.JitterNoiseLimit;
network.EndEpochEvent += new TrainingEpochEventHandler(
delegate(object senderNetwork, TrainingEpochEventArgs args)
{
// TODO: trainning error needs to be calculated
OnRunningEpoch(new AnnModelRunEpochEventArgs(args.TrainingIteration + 1, 0));
});
network.Learn(ForecastingDataSets.ConvertToTrainingSet(datasets), para.Iterations);
datasets.ForecastedData = new double[datasets.InputData.Length][];
for (int i = 0; i < datasets.InputData.Length; i++)
{
datasets.ForecastedData[i] = new double[1];
datasets.ForecastedData[i][0] = Forecast(datasets.InputData[i]);
}
OnFinishRunning(new ComponentRunEventArgs(datasets));
}
void CreateNewNetwork()
{
LinearLayer inputLayer = new LinearLayer(neurons);
SigmoidLayer hiddenLayer = new SigmoidLayer(hidden1Neurons);
SigmoidLayer hiddenLayer2 = new SigmoidLayer(hidden2Neurons);
LinearLayer outputLayer = new LinearLayer(outputNum);
BackpropagationConnector conn1 = new BackpropagationConnector(inputLayer, hiddenLayer);
conn1.Initializer = new RandomFunction(0d, 0.001d);
BackpropagationConnector conn3 = new BackpropagationConnector(hiddenLayer, hiddenLayer2);
conn3.Initializer = new RandomFunction(0d, 0.001d);
BackpropagationConnector conn2 = new BackpropagationConnector(hiddenLayer2, outputLayer);
conn2.Initializer = new RandomFunction(0d, 0.001d);
conn1.Initialize();
conn2.Initialize();
conn3.Initialize();
neuralNetwork = new BackpropagationNetwork(inputLayer, outputLayer);
neuralNetwork.SetLearningRate(learningRate);
neuralNetwork.Initialize();
}
public void AddOutboundConnection_ValidatesArgs()
{
// Setup
var network = new BackpropagationNetwork(new Mock<IErrorCalculator>().Object);
// Execute/Verify
network.AddOutboundConnection(null);
}
protected override void SolveInstance(IGH_DataAccess DA)
{
CrowNetBP net = new CrowNetBP();
if (!networkLoaded)
{
int cycles = 1000;
GH_Structure <GH_Number> tiv = new GH_Structure <GH_Number>();
GH_Structure <GH_Number> tov = new GH_Structure <GH_Number>();
DA.GetData(0, ref cycles);
DA.GetData(1, ref net);
DA.GetDataTree(2, out tiv);
DA.GetDataTree(3, out tov);
double[][] trainInVectors = Utils.GHTreeToMultidimensionalArray(tiv);
double[][] trainOutVectors = Utils.GHTreeToMultidimensionalArray(tov);
int trainVectorCount = trainInVectors.Length;
if (trainVectorCount != trainOutVectors.Length)
{
AddRuntimeMessage(GH_RuntimeMessageLevel.Error, "Please supply an equal amount of input and output training vectors!");
}
int trainInVectorDimension = trainInVectors[0].Length;
int trainOutVectorDimension = trainOutVectors[0].Length;
BackpropagationNetwork network = net.network(trainInVectorDimension, trainOutVectorDimension);
// set Trainingset
TrainingSet trainingSet = new TrainingSet(trainInVectorDimension, trainOutVectorDimension);
for (int i = 0; i < trainVectorCount; i++)
{
trainingSet.Add(new TrainingSample(trainInVectors[i], trainOutVectors[i]));
}
// train
network.Learn(trainingSet, cycles);
this.Network = network;
}
if (this.Network != null)
{
DA.SetData(0, this.Network.MeanSquaredError.ToString("0.0000000000"));
CrowNetBPP nn = new CrowNetBPP(this.Network);
nn.hiddenLayerList = net.hiddenLayerList;
nn.layerStructure = net.layerStructure;
nn.neuronCount = net.neuronCount;
DA.SetData(1, nn);
}
networkLoaded = false;
}
private void Form1_OnLoad(object sender, EventArgs e)
{
inputTier = new LinearLayer(35);
hiddenTier = new SigmoidLayer(3);
outputTier = new SigmoidLayer(5);
_ = new BackpropagationConnector(inputTier, hiddenTier);
_ = new BackpropagationConnector(hiddenTier, outputTier);
neuralNetwork = new BackpropagationNetwork(inputTier, outputTier);
neuralNetwork.Initialize();
}
private void WFAnnRecognition_Load(object sender, EventArgs e)
{
lstLog.Items.Insert(0, "Initialize ANN model");
inputLayer = new LinearLayer(35);
hiddenLayer = new SigmoidLayer(3);
outputLayer = new SigmoidLayer(5);
BackpropagationConnector connector = new BackpropagationConnector(inputLayer, hiddenLayer);
BackpropagationConnector connector2 = new BackpropagationConnector(hiddenLayer, outputLayer);
network = new BackpropagationNetwork(inputLayer, outputLayer);
network.Initialize();
}
/// <summary>
/// 点击计算按钮
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
private void tsmiCalculate_Click(object sender, EventArgs e)
{
// 创建输入层、隐层和输出层
ActivationLayer inputLayer = GetLayer(cboInputLayerType.SelectedItem.ToString(), 2);
ActivationLayer hiddenLayer = GetLayer(cboHiddenLayerType.SelectedItem.ToString(), int.Parse(txtHiddenLayerCount.Text));
ActivationLayer outputLayer = GetLayer(cboOutputLayerType.SelectedItem.ToString(), 1);
// 创建层之间的关联
new BackpropagationConnector(inputLayer, hiddenLayer, ConnectionMode.Complete).Initializer = new RandomFunction(0, 0.3);
new BackpropagationConnector(hiddenLayer, outputLayer, ConnectionMode.Complete).Initializer = new RandomFunction(0, 0.3);
// 创建神经网络
var network = new BackpropagationNetwork(inputLayer, outputLayer);
network.SetLearningRate(double.Parse(txtInitialLearningRate.Text), double.Parse(txtFinalLearningRate.Text));
// 进行训练
var trainingSet = new TrainingSet(2, 1);
for (var i = 0; i < 17; i++)
{
var x1 = data[i, 0];
var x2 = data[i, 1];
var y = data[i, 2];
var inputVector = new double[] { x1, x2 };
var outputVector = new double[] { y };
var trainingSample = new TrainingSample(inputVector, outputVector);
trainingSet.Add(trainingSample);
}
network.SetLearningRate(0.3, 0.1);
network.Learn(trainingSet, int.Parse(txtTrainingEpochs.Text));
network.StopLearning();
// 进行预测
for (var i = 0; i < 17; i++)
{
var x1 = data[i, 0];
var x2 = data[i, 1];
var y = data[i, 2];
var testInput = new double[] { x1, x2 };
var testOutput = network.Run(testInput)[0];
var absolute = testOutput - y;
var relative = Math.Abs((testOutput - y) / testOutput);
dgvData.Rows[i].Cells[3].Value = testOutput.ToString("f3");
dgvData.Rows[i].Cells[4].Value = absolute.ToString("f3");
dgvData.Rows[i].Cells[5].Value = (relative * 100).ToString("f1") + "%";
}
}
private void Form1_Load(object sender, EventArgs e)
{
int gizlikatmansayisi = Convert.ToInt32(Microsoft.VisualBasic.Interaction.InputBox("Gizli Katman Sayısını Giriniz", "Bilgi Girişi", "Örn: 3", 0, 0));
giriskatmanı = new LinearLayer(35);
gizlikatman = new SigmoidLayer(gizlikatmansayisi);
cikiskatmani = new SigmoidLayer(5);
BackpropagationConnector giris_gizli_baglanti = new BackpropagationConnector(giriskatmanı, gizlikatman);
BackpropagationConnector gizli_cikis_baglanti = new BackpropagationConnector(gizlikatman, cikiskatmani);
ag = new BackpropagationNetwork(giriskatmanı, cikiskatmani);
ag.Initialize();
}
public Imagine(int w, int h)
{
W = w;
H = h;
Inputs = W * H * 3;
iLay = new LinearLayer(Inputs);
hLay = new SigmoidLayer(neuronCount);
oLay = new SigmoidLayer(w * h * 3);
c1 = new BackpropagationConnector(iLay, hLay, ConnectionMode.Complete);
c2 = new BackpropagationConnector(hLay, oLay, ConnectionMode.Complete);
net = new BackpropagationNetwork(iLay, oLay);
net.SetLearningRate(learningRate);
}
private int CreateNeuralNetwork(int input, int output, int[] hidden)
{
LinearLayer inputLayer = new LinearLayer(input);
SigmoidLayer outputLayer = new SigmoidLayer(output);
// minimum size
if (hidden == null)
{
hidden = new int[] { input + 1, input + 1 };
}
var hiddenLayers = new SigmoidLayer[hidden.Length];
// plus two because of the input and the output layers
var connectors = new BackpropagationConnector[hidden.Length + 2];
// create the hidden layers
for (int k = 0; k < hidden.Length; k++)
{
hiddenLayers[k] = new SigmoidLayer(hidden[k]);
}
// back propagation from first hidden layer to input
connectors[0] = new BackpropagationConnector(inputLayer, hiddenLayers[0]);
// back propagation between the hidden layers
for (int k = 1; k < hidden.Length; k++)
{
connectors[k] = new BackpropagationConnector(hiddenLayers[k - 1], hiddenLayers[k]);
}
// back propagation from output to last hidden layer
connectors[hidden.Length - 1] = new BackpropagationConnector(hiddenLayers[hidden.Length - 1], outputLayer);
// The network
neuralNetwork = new BackpropagationNetwork(inputLayer, outputLayer);
#region retrieve network weight count
int netWeightCount = 0;
foreach (BackpropagationConnector connector in neuralNetwork.Connectors)
{
foreach (BackpropagationSynapse synapse in connector.Synapses)
{
netWeightCount += 2;
}
}
#endregion
return(netWeightCount);
}
public static void setNetworkWeights(BackpropagationNetwork aNetwork, double[] weights)
{
// Setup the network's weights.
int index = 0;
foreach (BackpropagationConnector connector in aNetwork.Connectors)
{
foreach (BackpropagationSynapse synapse in connector.Synapses)
{
synapse.Weight = weights[index++];
// synapse.SourceNeuron.SetBias(weights[index++]);
}
}
}
public void initNet(int fLayer, int sLayer, int tLayer)
{
ActivationLayer inputLayer = createLayer(fLayer, inputNeurons);
ActivationLayer hiddenLayer = createLayer(sLayer, hiddenNeurons);
ActivationLayer outputLayer = createLayer(tLayer, outputNeurons);
//var hiddenLayer = new LinearLayer(hiddenNeurons);
//var outputLayer = new LinearLayer(outputNeurons);
new BackpropagationConnector(inputLayer, hiddenLayer).Initializer = new RandomFunction(0d, 0.3d);
new BackpropagationConnector(hiddenLayer, outputLayer).Initializer = new RandomFunction(0d, 0.3d);
network = new BackpropagationNetwork(inputLayer, outputLayer);
network.SetLearningRate(learningRate);
SampleInput = new double[inputNeurons + outputNeurons];
SampleOutput = new double[outputNeurons];
}
/// <summary>
/// Sequential update the network from select layer with multitherading
/// </summary>
/// <param name="StartLayer">Start layer</param>
public static void SequentialUpdateMultitherading(this BackpropagationNetwork backpropagation, int StartLayer)
{
for (int layer = StartLayer; layer < backpropagation.Neurons.Count; layer++)
{
int neuronCount = backpropagation.Neurons[layer].Length;
IEnumerable <Neuron> prevLayer = (layer == 0) ? (backpropagation.InputNeurons) : (backpropagation.Neurons[layer - 1]);
Parallel.For(0, neuronCount, delegate(int i)
{
Neuron neuron = backpropagation.Neurons[layer][i];
neuron.Update(prevLayer.Select((Neuron n) => (n.OutputValue)).ToArray(), null, double.NaN);
});
}
}
private void button8_Click(object sender, EventArgs e)
{
LinearLayer inputLayer = new LinearLayer(Convert.ToInt32(textBox3.Text));
SigmoidLayer hiddenLayer = new SigmoidLayer(Convert.ToInt32(textBox4.Text));
SigmoidLayer outputLayer = new SigmoidLayer(Convert.ToInt32(textBox5.Text));
BackpropagationConnector conn1 = new BackpropagationConnector(inputLayer, hiddenLayer);
BackpropagationConnector conn2 = new BackpropagationConnector(hiddenLayer, outputLayer);
network = new BackpropagationNetwork(inputLayer, outputLayer);
network.Initialize();
MessageBox.Show("Rete generata con successo.");
}
async Task LearnNetworkAsync()
{
_network = new BackpropagationNetwork(_inputLayer, _outputLayer);
_network.Initialize();
var trainingSet = new TrainingSet(1, 3);
foreach (var b in from bomb in Enum.GetValues(typeof(BombTypes)).Cast <BombTypes>() where bomb != BombTypes.Mine select BombFabric.CreateBomb(bomb) into b where b != null select b)
{
trainingSet.Add(new TrainingSample(new double[] { b.BeepsLevel },
new double[] {
(int)b.FirstStageDisarming,
(int)b.SecondStageDisarming,
(int)b.ThirdStageDisarming
}));
}
_network.Learn(trainingSet, 100000);
}
public static void UpdatePositiveMultitherading(this BackpropagationNetwork backpropagation, double[] Input)
{
// Check parameters
if (Input.Length != backpropagation.InputNeurons.Length)
{
throw new ArgumentException("Invalid input array, its size is different from the number of input neurons.", nameof(Input));
}
// Update input neurons
Parallel.For(0, backpropagation.InputNeurons.Length, delegate(int i)
{
backpropagation.InputNeurons[i].Update(Input[i]);
});
// Sequential update neurons
backpropagation.SequentialUpdateMultitherading(0);
}
private void StopLearning(object sender, EventArgs e)
{
if (network != null)
{
network.StopLearning();
LineItem lineItem = new LineItem("Approximated Function");
for (double xVal = 0; xVal < 10; xVal += 0.05d)
{
lineItem.AddPoint(xVal, network.Run(new double[] { xVal })[0]);
}
lineItem.Symbol.Type = SymbolType.None;
lineItem.Color = Color.DarkOrchid;
functionGraph.GraphPane.CurveList.Add(lineItem);
functionGraph.Refresh();
functionGraph.GraphPane.CurveList.Remove(lineItem);
}
network = null;
EnableControls(true);
}
private void load_btn_Clicked(object sender, EventArgs e)
{
OpenFileDialog openFileDialog = new OpenFileDialog();
openFileDialog.Filter = "CNW files (*.cnw)|*.cnw";
openFileDialog.InitialDirectory = ".";
if (openFileDialog.ShowDialog() == DialogResult.OK)
{
try
{
System.IO.Stream stream = System.IO.File.Open(openFileDialog.FileName, System.IO.FileMode.Open);
BinaryFormatter bFormatter = new BinaryFormatter();
this.Network = bFormatter.Deserialize(stream) as BackpropagationNetwork;
stream.Close();
networkLoaded = true;
ExpireSolution(true);
}
catch (Exception ex) { AddRuntimeMessage(GH_RuntimeMessageLevel.Error, "Couldn't load the crow net:\n" + ex.Message); }
}
}
public void setNeuralNetwork(BackpropagationNetwork network)
{
this._neuralNetwork = network;
}
private void Start(object sender, EventArgs e)
{
CleanseGraph();
EnableControls(false);
curve.Color = enabledColor;
if (!int.TryParse(txtCycles.Text, out cycles))
{
cycles = 10000;
}
if (!double.TryParse(txtLearningRate.Text, out learningRate))
{
learningRate = 0.25d;
}
if (!int.TryParse(txtNeuronCount.Text, out neuronCount))
{
neuronCount = 10;
}
if (cycles <= 0)
{
cycles = 10000;
}
if (learningRate < 0 || learningRate > 1)
{
learningRate = 0.25d;
}
if (neuronCount <= 0)
{
neuronCount = 10;
}
txtCycles.Text = cycles.ToString();
txtLearningRate.Text = learningRate.ToString();
txtNeuronCount.Text = neuronCount.ToString();
LinearLayer inputLayer = new LinearLayer(1);
SigmoidLayer hiddenLayer = new SigmoidLayer(neuronCount);
SigmoidLayer outputLayer = new SigmoidLayer(1);
new BackpropagationConnector(inputLayer, hiddenLayer).Initializer = new RandomFunction(0d, 0.3d);
new BackpropagationConnector(hiddenLayer, outputLayer).Initializer = new RandomFunction(0d, 0.3d);
network = new BackpropagationNetwork(inputLayer, outputLayer);
network.SetLearningRate(learningRate);
TrainingSet trainingSet = new TrainingSet(1, 1);
for (int i = 0; i < curve.Points.Count; i++)
{
double xVal = curve.Points[i].X;
for (double input = xVal - 0.05; input < xVal + 0.06; input += 0.01)
{
trainingSet.Add(new TrainingSample(new double[] { input }, new double[] { curve.Points[i].Y }));
}
}
network.EndEpochEvent += new TrainingEpochEventHandler(
delegate(object senderNetwork, TrainingEpochEventArgs args)
{
trainingProgressBar.Value = (int)(args.TrainingIteration * 100d / cycles);
Application.DoEvents();
});
network.Learn(trainingSet, cycles);
StopLearning(this, EventArgs.Empty);
}
/// <summary>
/// This constructs a training procedure for standard backpropagation techniques.
/// More advanced ones will be used as seen in the example.
/// </summary>
/// <param name="writer"></param>
public TestingNdn(StreamWriter writer)
{
TrainingSample sample = new TrainingSample(
new double[] { },
new double[] { });
//We might make a gui for this later.
int numberOfNeurons = 3;
double learningRate = 0.5;
int numberOfCycles = 10000;
double[] errorList = new double[numberOfCycles];
LinearLayer inputLayer = new LinearLayer(2);
SigmoidLayer hiddenLayer = new SigmoidLayer(numberOfNeurons);
SigmoidLayer outputLayer = new SigmoidLayer(1);
// This layer is a event handler that fires when the output is generated, hence backpropagation.
BackpropagationConnector conn1 = new BackpropagationConnector(inputLayer, hiddenLayer);
BackpropagationConnector conn2 = new BackpropagationConnector(hiddenLayer, outputLayer);
BackpropagationNetwork network = new BackpropagationNetwork(inputLayer, outputLayer);
network.SetLearningRate(learningRate);
TrainingSet trainingSet = new TrainingSet(2, 1);
trainingSet.Add(new TrainingSample(new double[2] {
0, 0
}, new double[1] {
0
}));
trainingSet.Add(new TrainingSample(new double[2] {
0, 1
}, new double[1] {
1
}));
trainingSet.Add(new TrainingSample(new double[2] {
1, 0
}, new double[1] {
1
}));
trainingSet.Add(new TrainingSample(new double[2] {
1, 1
}, new double[1] {
0
}));
double max = 0;
// create an anonymous function to capture the error value of each iteration, and report back the percent of completion.
network.EndEpochEvent +=
delegate(object networkInput, TrainingEpochEventArgs args)
{
errorList[args.TrainingIteration] = network.MeanSquaredError;
max = Math.Max(max, network.MeanSquaredError);
PercentComplete = args.TrainingIteration * 100 / numberOfCycles;
};
network.Learn(trainingSet, numberOfCycles);
double[] indices = new double[numberOfCycles];
// for (int i = 0; i < numberOfCycles; i++) { indices[i] = i; } .. oh nvm, its for graphing the learning curve
// what to do for error list?
// errorList => for plotting stuff.
for (int i = 0; i < numberOfCycles; i++)
{
//Console.WriteLine(errorList[i]);
}
double[] outputResult = network.OutputLayer.GetOutput();
Console.WriteLine("final output");
double[] r1 = new double[] { 0, 0 };
double[] r2 = new double[] { 0, 1 };
double[] r3 = new double[] { 1, 0 };
double[] r4 = new double[] { 1, 1 };
Console.WriteLine(" 0 0 => " + network.Run(r1)[0]);
Console.WriteLine(" 0 1 => " + network.Run(r2)[0]);
Console.WriteLine(" 1 0 => " + network.Run(r3)[0]);
Console.WriteLine(" 1 1 => " + network.Run(r4)[0]);
}
public BackpropagationNetworkDeserializer(BackpropagationNetwork TargetNetwork)
{
targetNetwork = TargetNetwork;
}
public static void SaveTrainingSet(TrainingSet ts, string path, BackpropagationNetwork network)
{
StringBuilder sb = new StringBuilder();
using (StreamWriter sw = new StreamWriter(path))
{
double[] forecastVector;
// 写标题
for (int index = 1; index <= ts.InputVectorLength; index++)
{
sb.Append("Input" + index.ToString() + ",");
}
for (int index = 1; index <= ts.OutputVectorLength; index++)
{
sb.Append("Output" + index.ToString() + ",");
}
for (int index = 1; index <= ts.OutputVectorLength - 1; index++)
{
sb.Append("Forecast" + index.ToString() + ",");
}
sb.Append("Forecast" + ts.OutputVectorLength.ToString());
sw.WriteLine(sb.ToString());
// 重复样品
TrainingSample tsamp;
for (int row = 0; row < ts.TrainingSampleCount; row++)
{
tsamp = ts[row];
sb.Clear();
// 附加输入
for (int index = 0; index < ts.InputVectorLength; index++)
{
sb.Append(tsamp.InputVector[index].ToString("0.0000") + ",");
}
// 附加输出
for (int index = 0; index < ts.OutputVectorLength; index++)
{
sb.Append(tsamp.OutputVector[index].ToString("0.0000") + ",");
}
//
// 获取和追加预测
//
// 未标准化
forecastVector = network.Run(tsamp.InputVector);
// 标准化 - NOOOO
//forecastVector = network.Run(tsamp.NormalizedInputVector);
for (int index = 0; index < ts.OutputVectorLength - 1; index++)
{
sb.Append(forecastVector[index].ToString("0.0000") + ",");
}
sb.Append(forecastVector[ts.OutputVectorLength - 1].ToString("0.0000"));
// 输出这一行
sw.WriteLine(sb.ToString());
}
}
}
public void CalculateError_DistributesSignals()
{
//// SETUP
const double ErrorSignal1 = -0.5d;
const double ErrorSignal2 = 1.5d;
const double DigitizedErrorSignal1 = -1.0d;
const double DigitizedErrorSignal2 = 1.0d;
const double NetworkErrorSignal1 = -1.0d;
const double NetworkErrorSignal2 = 1.0d;
var expectedNetworkErrors = new[] { NetworkErrorSignal1, NetworkErrorSignal2 };
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound1 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound2 = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that all are reporting their respective errors
mockOutbound1.SetupGet(mock => mock.ErrorSignal).Returns(ErrorSignal1);
mockOutbound2.SetupGet(mock => mock.ErrorSignal).Returns(ErrorSignal2);
mockOutbound1.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(true);
// Create 2 input nodes and 5 output nodes.
var mockInputNode1 = new Mock<ISupervisedLearnerNode>();
var mockInputNode2 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode1 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode2 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode3 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode4 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode5 = new Mock<ISupervisedLearnerNode>();
// Program nodes to report input/output sizes.
mockInputNode1.SetupGet(mock => mock.InputSize).Returns(1);
mockInputNode2.SetupGet(mock => mock.InputSize).Returns(1);
mockOutputNode1.SetupGet(mock => mock.OutputSize).Returns(1);
mockOutputNode2.SetupGet(mock => mock.OutputSize).Returns(1);
mockOutputNode3.SetupGet(mock => mock.OutputSize).Returns(1);
mockOutputNode4.SetupGet(mock => mock.OutputSize).Returns(1);
mockOutputNode5.SetupGet(mock => mock.OutputSize).Returns(1);
// Program nodes to provide values.
mockInputNode1.SetupGet(mock => mock.CachedErrors).Returns(new[] { NetworkErrorSignal1 });
mockInputNode2.SetupGet(mock => mock.CachedErrors).Returns(new[] { NetworkErrorSignal2 });
mockOutputNode1.SetupGet(mock => mock.CachedOutputs).Returns(new[] { 0.0d });
mockOutputNode2.SetupGet(mock => mock.CachedOutputs).Returns(new[] { 0.0d });
mockOutputNode3.SetupGet(mock => mock.CachedOutputs).Returns(new[] { 0.0d });
mockOutputNode4.SetupGet(mock => mock.CachedOutputs).Returns(new[] { 0.0d });
mockOutputNode5.SetupGet(mock => mock.CachedOutputs).Returns(new[] { 0.0d });
// Create the test object.
var network = new BackpropagationNetwork(new Mock<IErrorCalculator>().Object);
network.AddInboundConnection(mockInbound1.Object);
network.AddInboundConnection(mockInbound2.Object);
network.AddOutboundConnection(mockOutbound1.Object);
network.AddOutboundConnection(mockOutbound2.Object);
network.AddInputNode(mockInputNode1.Object);
network.AddInputNode(mockInputNode2.Object);
network.AddOutputNode(mockOutputNode1.Object);
network.AddOutputNode(mockOutputNode2.Object);
network.AddOutputNode(mockOutputNode3.Object);
network.AddOutputNode(mockOutputNode4.Object);
network.AddOutputNode(mockOutputNode5.Object);
// EXECUTION
network.Fire(new[] { 0.0d });
network.CalculateError(ErrorSignal1);
// VERIFICATION: The IsReporting signals were checked...
mockOutbound1.Verify(mock => mock.IsReportingError, Times.Exactly(1));
mockOutbound2.Verify(mock => mock.IsReportingError, Times.Exactly(1));
// ...and calculation activities occurred.
mockInbound1.Verify(mock => mock.ReportError(NetworkErrorSignal1), Times.Once());
mockInbound2.Verify(mock => mock.ReportError(NetworkErrorSignal2), Times.Once());
mockOutbound1.VerifyGet(mock => mock.ErrorSignal, Times.Once());
mockOutbound1.Verify(mock => mock.ClearReportingFlag(), Times.Once());
mockOutbound2.VerifyGet(mock => mock.ErrorSignal, Times.Once());
mockOutbound2.Verify(mock => mock.ClearReportingFlag(), Times.Once());
mockInputNode1.VerifyGet(mock => mock.CachedErrors, Times.AtLeastOnce());
mockInputNode2.VerifyGet(mock => mock.CachedErrors, Times.AtLeastOnce());
mockOutputNode1.Verify(mock => mock.CalculateError(DigitizedErrorSignal1), Times.Once());
mockOutputNode2.Verify(mock => mock.CalculateError(DigitizedErrorSignal1), Times.Once());
mockOutputNode3.Verify(mock => mock.CalculateError(DigitizedErrorSignal1), Times.Once());
mockOutputNode4.Verify(mock => mock.CalculateError(DigitizedErrorSignal2), Times.Once());
mockOutputNode5.Verify(mock => mock.CalculateError(DigitizedErrorSignal2), Times.Once());
Assert.AreEqual(2, network.CachedErrors.Length);
Assert.AreEqual(expectedNetworkErrors, network.CachedErrors);
}
private void buttonOK_Click(object sender, EventArgs e)
{
this.Cursor = Cursors.WaitCursor;
try
{
//
// 解析数据文件
//
int inputCount = 0;
int outputCount = 1;
if (textInputCount.Text.Length > 0)
{
inputCount = int.Parse(textInputCount.Text);
}
if (textOutputCount.Text.Length > 0)
{
outputCount = int.Parse(textOutputCount.Text);
}
TrainingSet trainingSet = DataFile.CsvFileToTrainingSet(textTrainingSet.Text, ref inputCount, ref outputCount);
TrainingSet crossvalidSet = null;
if (textCvSet.Text.Length > 0)
{
crossvalidSet = DataFile.CsvFileToTrainingSet(textCvSet.Text, ref inputCount, ref outputCount);
}
//
// 创建新网络
//
// 输入层始终与输入计数是线性关系
LinearLayer inputLayer = new LinearLayer(inputCount);
// 创建隐层
ActivationLayer hiddenLayer1 = null;
ActivationLayer hiddenLayer2 = null;
ActivationLayer outputLayer = null;
if (comboActFunction1.SelectedIndex < 0)
{
MessageBox.Show("请选择激活函数！"); return;
}
switch ((HiddenLayerType)comboActFunction1.SelectedItem)
{
case HiddenLayerType.Linear: hiddenLayer1 = new LinearLayer(int.Parse(textNeuronCount1.Text)); break;
case HiddenLayerType.Logarithmic: hiddenLayer1 = new LogarithmLayer(int.Parse(textNeuronCount1.Text)); break;
case HiddenLayerType.Sigmoid: hiddenLayer1 = new SigmoidLayer(int.Parse(textNeuronCount1.Text)); break;
case HiddenLayerType.Sine: hiddenLayer1 = new SineLayer(int.Parse(textNeuronCount1.Text)); break;
case HiddenLayerType.Tanh: hiddenLayer1 = new TanhLayer(int.Parse(textNeuronCount1.Text)); break;
}
if (textNeuronCount2.Text.Length > 0 && int.Parse(textNeuronCount2.Text) > 0)
{
switch ((HiddenLayerType)comboActFunction2.SelectedItem)
{
case HiddenLayerType.Linear: hiddenLayer2 = new LinearLayer(int.Parse(textNeuronCount2.Text)); break;
case HiddenLayerType.Logarithmic: hiddenLayer2 = new LogarithmLayer(int.Parse(textNeuronCount2.Text)); break;
case HiddenLayerType.Sigmoid: hiddenLayer2 = new SigmoidLayer(int.Parse(textNeuronCount2.Text)); break;
case HiddenLayerType.Sine: hiddenLayer2 = new SineLayer(int.Parse(textNeuronCount2.Text)); break;
case HiddenLayerType.Tanh: hiddenLayer2 = new TanhLayer(int.Parse(textNeuronCount2.Text)); break;
}
}
if (comboOutputFunction.SelectedIndex < 0)
{
MessageBox.Show("请选择输出函数！"); return;
}
switch ((HiddenLayerType)comboOutputFunction.SelectedItem)
{
case HiddenLayerType.Linear: outputLayer = new LinearLayer(outputCount); break;
case HiddenLayerType.Logarithmic: outputLayer = new LogarithmLayer(outputCount); break;
case HiddenLayerType.Sigmoid: outputLayer = new SigmoidLayer(outputCount); break;
case HiddenLayerType.Sine: outputLayer = new SineLayer(outputCount); break;
case HiddenLayerType.Tanh: outputLayer = new TanhLayer(outputCount); break;
}
// 连接层, hidden2是可选的
new BackpropagationConnector(inputLayer, hiddenLayer1);
if (hiddenLayer2 != null)
{
new BackpropagationConnector(hiddenLayer1, hiddenLayer2);
new BackpropagationConnector(hiddenLayer2, outputLayer);
}
else
{
new BackpropagationConnector(hiddenLayer1, outputLayer);
}
BackpropagationNetwork backpropNetwork = new BackpropagationNetwork(inputLayer, outputLayer);
//
// 设置学习和退出参数
//
double startLearningRate = double.Parse(textStartLearningRate.Text);
double?finalLearningRate = null;
if (textFinalLearningRate.Text.Length > 0)
{
finalLearningRate = double.Parse(textFinalLearningRate.Text);
}
// 如果选择了学习率函数则使用
LearningRateFunction?lrf = null;
if (comboLRFunction.SelectedIndex > 0)
{
lrf = (LearningRateFunction)comboLRFunction.SelectedItem;
backpropNetwork.SetLearningRate(
LearningRateFactory.GetLearningRateFunction(lrf.Value, startLearningRate, finalLearningRate.Value));
}
else
{
// 否则使用普通学习率，也许有起点和终点
if (finalLearningRate.HasValue)
{
backpropNetwork.SetLearningRate(startLearningRate, finalLearningRate.Value);
}
else
{
backpropNetwork.SetLearningRate(startLearningRate);
}
}
// 如果给定，在连接器中设置动量
double?momentum = null;
if (textMomentum.Text.Length > 0)
{
momentum = double.Parse(textMomentum.Text);
foreach (ILayer layer in backpropNetwork.Layers)
{
foreach (BackpropagationConnector conn in layer.SourceConnectors)
{
conn.Momentum = momentum.Value;
}
foreach (BackpropagationConnector conn in layer.TargetConnectors)
{
conn.Momentum = momentum.Value;
}
}
}
//
// 新建工程和保存工程
//
int tmpInt;
NewProject = new NnProject();
NewProject.Network = backpropNetwork;
// 确保为新网络初始化权重，默认情况下新的训练周期将不会初始化
NewProject.Network.Initialize();
NewProject.ProjectName = textProjectName.Text.Trim();
NewProject.SaveFolder = textSaveFolder.Text;
NewProject.TrainingSet = trainingSet;
NewProject.CrossValidationSet = crossvalidSet;
NewProject.LearningParameters = new NnProject.NnLearningParameters();
NewProject.LearningParameters.InitialLearningRate = startLearningRate;
NewProject.LearningParameters.FinalLearningRate = finalLearningRate;
NewProject.LearningParameters.LearningRateFunction = lrf;
NewProject.LearningParameters.Momentum = momentum;
if (int.TryParse(textTrainingCycles.Text, out tmpInt))
{
NewProject.LearningParameters.MaxTrainingCycles = tmpInt;
}
NnProject.Save(NewProject, textSaveFolder.Text);
this.Close();
}
catch (Exception ex)
{
this.Cursor = Cursors.Default;
MessageBox.Show("Error creating network - " + ex.Message, "Error", MessageBoxButtons.OK, MessageBoxIcon.Error);
}
finally
{ this.Cursor = Cursors.Default; }
}
private void Train(object sender, EventArgs e)
{
// btnTrain.Enabled = false;
int cycles = 200;
// if (!int.TryParse(txtCycles.Text, out cycles)) { cycles = 200; }
// txtCycles.Text = cycles.ToString();
int currentCombination = 0;
//int totalCombinations = Alphabet.LetterCount * (Alphabet.LetterCount - 1) / 2;
for (int i = 0; i < Alphabet.LetterCount; i++)
{
for (int j = i + 1; j < Alphabet.LetterCount; j++)
{
ActivationLayer inputLayer = new LinearLayer(400);
ActivationLayer hiddenLayer = new SigmoidLayer(4);
ActivationLayer outputLayer = new SigmoidLayer(2);
new BackpropagationConnector(inputLayer, hiddenLayer);
new BackpropagationConnector(hiddenLayer, outputLayer);
BackpropagationNetwork network = new BackpropagationNetwork(inputLayer, outputLayer);
TrainingSet trainingSet = new TrainingSet(400, 2);
Alphabet ithLetter = Alphabet.GetLetter(i);
Alphabet jthLetter = Alphabet.GetLetter(j);
foreach (Letter instance in ithLetter.Instances)
{
trainingSet.Add(new TrainingSample(instance.GetEquivalentVector(20, 20), new double[] { 1d, 0d }));
}
foreach (Letter instance in jthLetter.Instances)
{
trainingSet.Add(new TrainingSample(instance.GetEquivalentVector(20, 20), new double[] { 0d, 1d }));
}
//progressTraining.Value = 100 * currentCombination / totalCombinations;
Application.DoEvents();
bool correct = false;
int currentCycles = 35;
int count = trainingSet.TrainingSampleCount;
while (correct == false & currentCycles <= cycles)
{
network.Initialize();
network.Learn(trainingSet, currentCycles);
correct = true;
for (int sampleIndex = 0; sampleIndex < count; sampleIndex++)
{
double[] op = network.Run(trainingSet[sampleIndex].InputVector);
if (((trainingSet[sampleIndex].OutputVector[0] > trainingSet[sampleIndex].OutputVector[1]) && op[0] - op[1] < 0.4) || ((trainingSet[sampleIndex].OutputVector[0] < trainingSet[sampleIndex].OutputVector[1]) && op[1] - op[0] < 0.4))
{
correct = false;
trainingSet.Add(trainingSet[sampleIndex]);
}
}
currentCycles *= 2;
}
//lstLog.Items.Add(cboAplhabet.Items[i] + " & " + cboAplhabet.Items[j] + " = " + network.MeanSquaredError.ToString("0.0000"));
// lstLog.TopIndex = lstLog.Items.Count - (int)(lstLog.Height / lstLog.ItemHeight);
try
{
using (Stream stream = File.Open(Application.StartupPath + @"\Networks\" + i.ToString("00") + j.ToString("00") + ".ndn", FileMode.Create))
{
IFormatter formatter = new BinaryFormatter();
formatter.Serialize(stream, network);
}
}
catch (Exception)
{
MessageBox.Show("Failed to save trained neural networks", "Critical Error", MessageBoxButtons.OK, MessageBoxIcon.Error);
return;
}
currentCombination++;
}
}
// progressTraining.Value = 0;
// btnTrain.Enabled = false;
}
static void Main(string[] args)
{
Boolean finished = false;
//Layer
LinearLayer inputLayer = new LinearLayer(25);
SigmoidLayer hiddenLayer1 = new SigmoidLayer(100);
SigmoidLayer outputLayer = new SigmoidLayer(4);
//Connectors
BackpropagationConnector connector = new BackpropagationConnector(inputLayer, hiddenLayer1);
BackpropagationConnector connector3 = new BackpropagationConnector(hiddenLayer1, outputLayer);
network = new BackpropagationNetwork(inputLayer, outputLayer);
network.Initialize();
labyrinth laby = new labyrinth();
player plyr = new player();
laby.Update(laby);
Console.WriteLine(laby.GetPlayerPosition(laby.Laby));
do
{
string pressedKey = Console.ReadKey(true).Key.ToString();
switch (pressedKey)
{
case "RightArrow":
if (laby.Laby[plyr.getPlayerPosition().Item1, plyr.getPlayerPosition().Item2 + 1] != laby.Wall)
{
laby.Laby = laby.SetPosition(plyr.getPlayerPosition(), plyr.move("R"), laby.Laby);
laby.Update(laby);
}
break;
case "LeftArrow":
if (laby.Laby[plyr.getPlayerPosition().Item1, plyr.getPlayerPosition().Item2 - 1] != laby.Wall)
{
laby.Laby = laby.SetPosition(plyr.getPlayerPosition(), plyr.move("L"), laby.Laby);
laby.Update(laby);
}
break;
case "UpArrow":
if (laby.Laby[plyr.getPlayerPosition().Item1 - 1, plyr.getPlayerPosition().Item2] != laby.Wall)
{
laby.Laby = laby.SetPosition(plyr.getPlayerPosition(), plyr.move("U"), laby.Laby);
laby.Update(laby);
}
break;
case "DownArrow":
if (laby.Laby[plyr.getPlayerPosition().Item1 + 1, plyr.getPlayerPosition().Item2] != laby.Wall)
{
laby.Laby = laby.SetPosition(plyr.getPlayerPosition(), plyr.move("D"), laby.Laby);
laby.Update(laby);
}
break;
}
} while (finished != true);
Console.Read();
}
public void ApplyWeightAdjustments_PropogatesThroughNetwork()
{
//// SETUP
const double ErrorSignal = -2.3d;
const float Momentum = 0.9f;
const float LearningRate = 0.1f;
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound1 = new Mock<ISupervisedLearnerConnection>();
var mockOutbound2 = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that both are reporting error signals
mockOutbound1.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutbound2.SetupGet(mock => mock.IsReportingError).Returns(true);
// Create 2 input nodes and 2 output nodes.
var mockInputNode1 = new Mock<ISupervisedLearnerNode>();
var mockInputNode2 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode1 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode2 = new Mock<ISupervisedLearnerNode>();
// Program nodes to report input/output sizes.
mockInputNode1.SetupGet(mock => mock.InputSize).Returns(1);
mockInputNode2.SetupGet(mock => mock.InputSize).Returns(1);
mockOutputNode1.SetupGet(mock => mock.OutputSize).Returns(1);
mockOutputNode2.SetupGet(mock => mock.OutputSize).Returns(1);
// Program nodes to provide values.
mockInputNode1.SetupGet(mock => mock.CachedErrors).Returns(new[] { ErrorSignal });
mockInputNode2.SetupGet(mock => mock.CachedErrors).Returns(new[] { ErrorSignal });
mockOutputNode1.SetupGet(mock => mock.CachedOutputs).Returns(new[] { 0.0d });
mockOutputNode2.SetupGet(mock => mock.CachedOutputs).Returns(new[] { 0.0d });
// Create the test object.
var network = new BackpropagationNetwork(new Mock<IErrorCalculator>().Object);
network.AddInboundConnection(mockInbound1.Object);
network.AddInboundConnection(mockInbound2.Object);
network.AddOutboundConnection(mockOutbound1.Object);
network.AddOutboundConnection(mockOutbound2.Object);
network.AddInputNode(mockInputNode1.Object);
network.AddInputNode(mockInputNode2.Object);
network.AddOutputNode(mockOutputNode1.Object);
network.AddOutputNode(mockOutputNode2.Object);
network.Fire(new[] { 0.0d });
network.CalculateError(ErrorSignal);
// EXECUTION
network.ApplyWeightAdjustments(LearningRate, Momentum);
// VERIFICATION
mockInputNode1.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Once());
mockInputNode2.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Once());
mockOutbound1.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Once());
mockOutbound2.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Once());
}
private void RefreshScreenForCurrentNetWork()
{
txtCycles.Text = currentProject.LearningParameters.MaxTrainingCycles.ToString();
txtInitialLearningRate.Text = currentProject.LearningParameters.InitialLearningRate.ToString();
if (currentProject.LearningParameters.FinalLearningRate.HasValue)
{
txtFinalLearningRate.Text = currentProject.LearningParameters.FinalLearningRate.ToString();
}
else
{
txtFinalLearningRate.Text = "";
}
if (currentProject.LearningParameters.LearningRateFunction.HasValue)
{
comboLRFunction.SelectedIndex = (int)currentProject.LearningParameters.LearningRateFunction + 1;
}
if (currentProject.LearningParameters.Momentum.HasValue)
{
textMomentum.Text = currentProject.LearningParameters.Momentum.ToString();
}
else
{
textMomentum.Text = "";
}
string networkDesc = "";
BackpropagationNetwork bn = currentProject.Network;
int layerNo = 1;
foreach (ILayer layer in bn.Layers)
{
networkDesc += "Layer " + layerNo.ToString() + ": {" + layer.ToString().Replace("NeuronDotNet.Core.Backpropagation.", "") + "," + layer.NeuronCount.ToString() + "}" + System.Environment.NewLine;
layerNo++;
}
networkDesc += System.Environment.NewLine + "TrainingSampleCount: " + currentProject.TrainingSet.TrainingSampleCount.ToString();
networkDescription.Text = networkDesc;
lblTrainErrorVal.Text = bn.MeanSquaredError.ToString("0.000000");
if (currentProject.CrossValidationSet != null)
{
double cvError = GetCrossValidationError();
lblSumSqErrorCV.Text = cvError.ToString("0.000000");
double percErr = Math.Abs(100 * (currentProject.Network.MeanSquaredError - cvError) / currentProject.Network.MeanSquaredError);
if (!double.IsInfinity(percErr))
{
lblCvPercError.Text = percErr.ToString("0.00") + " %";
}
else
{
lblCvPercError.Text = "";
}
lblSumSqErrorCV.Enabled = true;
lblSumSqErrorCV.BackColor = lblTrainErrorVal.BackColor;
}
else
{
lblSumSqErrorCV.Text = ""; lblCvPercError.Text = "";
lblSumSqErrorCV.Enabled = false;
lblSumSqErrorCV.BackColor = lblCVErrLabel.BackColor;
}
}
public void Train_TrainsTheNetwork()
{
//// SETUP
const int NumEpochs = 3;
const float LearningRate = 0.1f;
const float Momentum = 0.9f;
const double FinalError = 99.99d;
var inputs = new[] { new[] { 1.0d, 2.0d }, new[] { 3.0d, 4.0d } };
var ideals = new[] { new[] { 5.0d, 6.0d }, new[] { 7.0d, 8.0d } };
var outputs = new[] { new[] { 3.5d, 5.5d }, new[] { 7.5d, 9.5d } };
var overallErrors = new[]
{
new[] { ideals[0][0] - outputs[0][0], ideals[0][1] - outputs[0][1] },
new[] { ideals[1][0] - outputs[1][0], ideals[1][1] - outputs[1][1] }
};
var finalErrors = new[] { new[] { 1.5d, 2.5d }, new[] { 3.5d, 4.5d } };
var mockErrorCalc = new Mock<IErrorCalculator>();
mockErrorCalc.Setup(mock => mock.Calculate()).Returns(FinalError);
// Create 2 input nodes and 2 output nodes.
var mockInputNode1 = new Mock<ISupervisedLearnerNode>();
var mockInputNode2 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode1 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode2 = new Mock<ISupervisedLearnerNode>();
// Program nodes to report input/output sizes.
mockInputNode1.SetupGet(mock => mock.InputSize).Returns(1);
mockInputNode2.SetupGet(mock => mock.InputSize).Returns(1);
mockOutputNode1.SetupGet(mock => mock.OutputSize).Returns(1);
mockOutputNode2.SetupGet(mock => mock.OutputSize).Returns(1);
// Program output nodes to return output values based on inputs.
mockInputNode1.Setup(mock => mock.Fire(new[] { inputs[0][0] })).Callback(() => mockOutputNode1.SetupGet(mock => mock.CachedOutputs).Returns(new[] { outputs[0][0] }));
mockInputNode2.Setup(mock => mock.Fire(new[] { inputs[0][1] })).Callback(() => mockOutputNode2.SetupGet(mock => mock.CachedOutputs).Returns(new[] { outputs[0][1] }));
mockInputNode1.Setup(mock => mock.Fire(new[] { inputs[1][0] })).Callback(() => mockOutputNode1.SetupGet(mock => mock.CachedOutputs).Returns(new[] { outputs[1][0] }));
mockInputNode2.Setup(mock => mock.Fire(new[] { inputs[1][1] })).Callback(() => mockOutputNode2.SetupGet(mock => mock.CachedOutputs).Returns(new[] { outputs[1][1] }));
// Program input nodes to return final error based on overal errors.
mockOutputNode1.Setup(mock => mock.CalculateError(overallErrors[0][0])).Callback(
() => mockInputNode1.SetupGet(mock => mock.CachedErrors).Returns(new[] { finalErrors[0][0] }));
mockOutputNode1.Setup(mock => mock.CalculateError(overallErrors[1][0])).Callback(
() => mockInputNode1.SetupGet(mock => mock.CachedErrors).Returns(new[] { finalErrors[1][0] }));
mockOutputNode2.Setup(mock => mock.CalculateError(overallErrors[0][1])).Callback(
() => mockInputNode2.SetupGet(mock => mock.CachedErrors).Returns(new[] { finalErrors[0][1] }));
mockOutputNode2.Setup(mock => mock.CalculateError(overallErrors[1][1])).Callback(
() => mockInputNode2.SetupGet(mock => mock.CachedErrors).Returns(new[] { finalErrors[1][1] }));
// Create the network
var network = new BackpropagationNetwork(mockErrorCalc.Object);
network.AddInputNode(mockInputNode1.Object);
network.AddInputNode(mockInputNode2.Object);
network.AddOutputNode(mockOutputNode1.Object);
network.AddOutputNode(mockOutputNode2.Object);
// Execute
var actualError = network.Train(NumEpochs, LearningRate, Momentum, inputs, ideals);
//// Verify
// Use of error calculator
mockErrorCalc.Verify(mock => mock.Reset(), Times.Exactly(NumEpochs));
mockErrorCalc.Verify(mock => mock.AddToErrorCalc(overallErrors[0]), Times.Exactly(NumEpochs));
mockErrorCalc.Verify(mock => mock.AddToErrorCalc(overallErrors[1]), Times.Exactly(NumEpochs));
mockErrorCalc.Verify(mock => mock.Calculate(), Times.Once());
// Use of input nodes
mockInputNode1.Verify(mock => mock.Fire(new[] { inputs[0][0] }), Times.Exactly(NumEpochs));
mockInputNode1.Verify(mock => mock.Fire(new[] { inputs[1][0] }), Times.Exactly(NumEpochs));
mockInputNode2.Verify(mock => mock.Fire(new[] { inputs[0][1] }), Times.Exactly(NumEpochs));
mockInputNode2.Verify(mock => mock.Fire(new[] { inputs[1][1] }), Times.Exactly(NumEpochs));
mockInputNode1.Verify(mock => mock.ClearCachedErrors(), Times.Exactly(NumEpochs * inputs.Length));
mockInputNode2.Verify(mock => mock.ClearCachedErrors(), Times.Exactly(NumEpochs * inputs.Length));
mockInputNode1.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Exactly(NumEpochs));
mockInputNode2.Verify(mock => mock.ApplyWeightAdjustments(LearningRate, Momentum), Times.Exactly(NumEpochs));
// Use of output nodes
mockOutputNode1.VerifyGet(mock => mock.CachedOutputs, Times.AtLeast(NumEpochs));
mockOutputNode2.VerifyGet(mock => mock.CachedOutputs, Times.AtLeast(NumEpochs));
mockOutputNode1.Verify(mock => mock.CalculateError(overallErrors[0][0]), Times.Exactly(NumEpochs));
mockOutputNode1.Verify(mock => mock.CalculateError(overallErrors[1][0]), Times.Exactly(NumEpochs));
mockOutputNode2.Verify(mock => mock.CalculateError(overallErrors[0][1]), Times.Exactly(NumEpochs));
mockOutputNode2.Verify(mock => mock.CalculateError(overallErrors[1][1]), Times.Exactly(NumEpochs));
// Verify Error
Assert.AreEqual(FinalError, actualError);
}
public void CalculateError_DoesNotCalculateUntilAllConnectionsReport()
{
//// SETUP
// Create 2 inbound and 2 outbound mock connections.
var mockInbound1 = new Mock<ISupervisedLearnerConnection>();
var mockInbound2 = new Mock<ISupervisedLearnerConnection>();
var mockOutboundReporting = new Mock<ISupervisedLearnerConnection>();
var mockOutboundNotReporting = new Mock<ISupervisedLearnerConnection>();
// program the mock outbounds such that one is reporting and one isn't
mockOutboundReporting.SetupGet(mock => mock.IsReportingError).Returns(true);
mockOutboundNotReporting.SetupGet(mock => mock.IsReportingError).Returns(false);
// Create 2 input nodes and 2 output nodes.
var mockInputNode1 = new Mock<ISupervisedLearnerNode>();
var mockInputNode2 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode1 = new Mock<ISupervisedLearnerNode>();
var mockOutputNode2 = new Mock<ISupervisedLearnerNode>();
// Create the test object.
var network = new BackpropagationNetwork(new Mock<IErrorCalculator>().Object);
network.AddInboundConnection(mockInbound1.Object);
network.AddInboundConnection(mockInbound2.Object);
network.AddOutboundConnection(mockOutboundReporting.Object);
network.AddOutboundConnection(mockOutboundNotReporting.Object);
network.AddInputNode(mockInputNode1.Object);
network.AddInputNode(mockInputNode2.Object);
network.AddOutputNode(mockOutputNode1.Object);
network.AddOutputNode(mockOutputNode2.Object);
// EXECUTION
const double ErrorSignal = -2.3d;
network.CalculateError(ErrorSignal);
// VERIFICATION: The IsReporting signals were checked...
mockOutboundReporting.Verify(mock => mock.IsReportingError, Times.Exactly(1));
mockOutboundNotReporting.Verify(mock => mock.IsReportingError, Times.Exactly(1));
// ...but no calculation activities occurred.
mockInbound1.Verify(mock => mock.ReportError(It.IsAny<double>()), Times.Never());
mockInbound2.Verify(mock => mock.ReportError(It.IsAny<double>()), Times.Never());
mockOutboundReporting.Verify(mock => mock.ClearReportingFlag(), Times.Never());
mockOutboundNotReporting.Verify(mock => mock.ClearReportingFlag(), Times.Never());
mockInputNode1.VerifyGet(mock => mock.CachedErrors, Times.Never());
mockInputNode2.VerifyGet(mock => mock.CachedErrors, Times.Never());
mockOutputNode1.Verify(mock => mock.CalculateError(It.IsAny<double>()), Times.Never());
mockOutputNode2.Verify(mock => mock.CalculateError(It.IsAny<double>()), Times.Never());
}