/// <summary>
/// Override the Epoch() method
/// </summary>
/// <param name="trainingSet"></param>
/// <param name="modalitiesMeanSquarredError"></param>
/// <param name="globalMeanSquarred"></param>
public override void Epoch(List <Dictionary <Signal, double[, ]> > trainingSet, out Dictionary <Signal, double> modalitiesMeanSquarredError, out double globalMeanSquarred)
{
//Convert to Accord format
double[][] samples = new double[trainingSet.Count][];
for (int i = 0; i < trainingSet.Count; i++)
{
samples[i] = concatenateTrainingSample(trainingSet[i]);
}
for (int _layerIndex = 0; _layerIndex < network.Layers.Length; _layerIndex++)
{
//Create the teacher for the layer
teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = learningRate,
Momentum = momentum,
Decay = weightDecay,
},
LayerIndex = _layerIndex,
};
//Run the epoch for the layer
double[][] sampleForLayer = teacher.GetLayerInput(samples);
teacher.RunEpoch(sampleForLayer);
}
network.UpdateVisibleWeights();
//Run manually a base class epoch to have the exact same error measurement as other algos
learningLocked = true;
base.Epoch(trainingSet, out modalitiesMeanSquarredError, out globalMeanSquarred);
learningLocked = false;
}
private void learnLayerUnsupervised()
{
if (!Main.CanGenerate)
{
return;
}
Dispatcher dispatcher = Dispatcher.CurrentDispatcher;
new Task(() =>
{
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(Main.Network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = LearningRate,
Momentum = 0.5,
Decay = WeightDecay,
},
LayerIndex = SelectedLayerIndex - 1,
};
double[][] inputs;
Main.Database.Training.GetInstances(out inputs);
int batchCount = Math.Max(1, inputs.Length / BatchSize);
// Create mini-batches to speed learning
int[] groups = Accord.Statistics.Tools
.RandomGroups(inputs.Length, batchCount);
double[][][] batches = inputs.Subgroups(groups);
// Gather learning data for the layer
double[][][] layerData = teacher.GetLayerInput(batches);
var cd = teacher.GetLayerAlgorithm(teacher.LayerIndex) as ContrastiveDivergenceLearning;
// Start running the learning procedure
for (int i = 0; i < Epochs && !shouldStop; i++)
{
double error = teacher.RunEpoch(layerData) / inputs.Length;
dispatcher.BeginInvoke((Action <int, double>)updateError,
DispatcherPriority.ContextIdle, i + 1, error);
if (i == 10)
{
cd.Momentum = Momentum;
}
}
IsLearning = false;
}).Start();
}
private static DeepBeliefNetwork createNetwork(double[][] inputs)
{
DeepBeliefNetwork network = new DeepBeliefNetwork(6, 2, 1);
network.Machines[0].Hidden.Neurons[0].Weights[0] = 0.00461421;
network.Machines[0].Hidden.Neurons[0].Weights[1] = 0.04337112;
network.Machines[0].Hidden.Neurons[0].Weights[2] = -0.10839599;
network.Machines[0].Hidden.Neurons[0].Weights[3] = -0.06234004;
network.Machines[0].Hidden.Neurons[0].Weights[4] = -0.03017057;
network.Machines[0].Hidden.Neurons[0].Weights[5] = 0.09520391;
network.Machines[0].Hidden.Neurons[0].Threshold = 0;
network.Machines[0].Hidden.Neurons[1].Weights[0] = 0.08263872;
network.Machines[0].Hidden.Neurons[1].Weights[1] = -0.118437;
network.Machines[0].Hidden.Neurons[1].Weights[2] = -0.21710971;
network.Machines[0].Hidden.Neurons[1].Weights[3] = 0.02332903;
network.Machines[0].Hidden.Neurons[1].Weights[4] = 0.00953116;
network.Machines[0].Hidden.Neurons[1].Weights[5] = 0.09870652;
network.Machines[0].Hidden.Neurons[1].Threshold = 0;
network.Machines[0].Visible.Neurons[0].Threshold = 0;
network.Machines[0].Visible.Neurons[1].Threshold = 0;
network.Machines[0].Visible.Neurons[2].Threshold = 0;
network.Machines[0].Visible.Neurons[3].Threshold = 0;
network.Machines[0].Visible.Neurons[4].Threshold = 0;
network.Machines[0].Visible.Neurons[5].Threshold = 0;
network.UpdateVisibleWeights();
DeepBeliefNetworkLearning target = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
};
for (int layer = 0; layer < 2; layer++)
{
target.LayerIndex = layer;
double[][] layerInputs = target.GetLayerInput(inputs);
int iterations = 5000;
double[] errors = new double[iterations];
for (int i = 0; i < iterations; i++)
{
errors[i] = target.RunEpoch(layerInputs);
}
}
return(network);
}
public void Train(double[][] i, double[][] o = null, int outputLength = 10, int hiddenLayer = -1)
{
if (n == null)
{
if (File.Exists(p))
{
n = DeepBeliefNetwork.Load(p);
}
else
{
outputLength = (o == null) ? outputLength : o[0].Length;
hiddenLayer = (hiddenLayer == -1) ? (int)Math.Log(i[0].Length, outputLength) : hiddenLayer;
List <int> layers = new List <int>();
for (int j = 0; j < hiddenLayer; j++)
{
layers.Add(i[0].Length);
}
layers.Add(outputLength);
n = new DeepBeliefNetwork(new BernoulliFunction(), i[0].Length, layers.ToArray());
new GaussianWeights(n).Randomize();
}
}
dynamic t;
if (o == null)
{
t = new DeepBeliefNetworkLearning(n)
{
Algorithm = (h, v, j) => new ContrastiveDivergenceLearning(h, v), LayerIndex = n.Machines.Count - 1,
};
while (true)
{
e = t.RunEpoch(t.GetLayerInput(i));
}
}
else
{
t = new DeepNeuralNetworkLearning(n)
{
Algorithm = (ann, j) => new ParallelResilientBackpropagationLearning(ann), LayerIndex = n.Machines.Count - 1,
};
while (true)
{
e = t.RunEpoch(t.GetLayerInput(i), o);
}
}
}
public void RunTraining1(Training1Parameters parameters)
{
LogInfoUsingBothLoggers("Started unsupervised training.");
var teacher = new DeepBeliefNetworkLearning(NeuralNetwork)
{
Algorithm = (hiddenLayer, visibleLayer, i) => new ContrastiveDivergenceLearning(hiddenLayer, visibleLayer)
{
LearningRate = parameters.LearningRate,
Momentum = parameters.Momentum,
Decay = parameters.Decay,
}
};
var inputs = _configuration.InputsOutputsData.Inputs;
// Setup batches of input for learning.
var batchCount = Math.Max(1, inputs.Length / 100);
// Create mini-batches to speed learning.
var groups = Accord.Statistics.Tools.RandomGroups(inputs.Length, batchCount);
var batches = inputs.Subgroups(groups);
// Unsupervised learning on each hidden layer, except for the output layer.
var guiLogIntensity = GetGuiLogIntensity(parameters.UnsupervisedEpochs);
for (int layerIndex = 0; layerIndex < NeuralNetwork.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
var layerData = teacher.GetLayerInput(batches);
foreach (int i in Enumerable.Range(1, parameters.UnsupervisedEpochs))
{
var error = teacher.RunEpoch(layerData) / inputs.Length;
var message = $"Layer: {layerIndex} Epoch: {i}, Error: {error}";
LogCurrentEpochResult(message, guiLogIntensity, i, parameters.UnsupervisedEpochs);
if (_skipPhaseRequest.RequestedAndUnhandled)
{
LogPhaseSkippnigAndNotifyHandled(i, parameters.UnsupervisedEpochs);
break;
}
}
}
}
private void TrainUnsupervised(double[][] batchInputs, int iterations, Action <double, int, string> progressCallback)
{
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
unsuperVisedTeacher.LayerIndex = layerIndex;
var layerData = unsuperVisedTeacher.GetLayerInput(batchInputs);
foreach (int i in Enumerable.Range(1, iterations))
{
var error = unsuperVisedTeacher.RunEpoch(layerData) / batchInputs.Length;
if (progressCallback != null)
{
progressCallback(error, i, $"Unsupervised Layer {layerIndex}");
}
if (this.ShouldStopTraning)
{
this.ShouldStopTraning = false;
break;
}
}
}
}
public static void Learn(double[][] inputs, double[][] outputs)
{
var n = (int)(count * 0.8);
var testInputs = inputs.Skip(n).ToArray();
var testOutputs = outputs.Skip(n).ToArray();
inputs = inputs.Take(n).ToArray();
outputs = outputs.Take(n).ToArray();
var network = new DeepBeliefNetwork(inputs.First().Length, 10, 10);
new GaussianWeights(network, 0.1).Randomize();
network.UpdateVisibleWeights();
// Setup the learning algorithm.
var teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Setup batches of input for learning.
int batchCount = Math.Max(1, inputs.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Classes.Random(inputs.Length, batchCount);
double[][][] batches = inputs.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
// Unsupervised learning on each hidden layer, except for the output layer.
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < 200; i++)
{
double error = teacher.RunEpoch(layerData) / inputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
}
// Supervised learning on entire network, to provide output classification.
var teacher2 = new BackPropagationLearning(network)
{
LearningRate = 0.1,
Momentum = 0.5
};
// Run supervised learning.
for (int i = 0; i < n; i++)
{
double error = teacher2.RunEpoch(inputs, outputs) / inputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
// Test the resulting accuracy.
int correct = 0;
for (int i = 0; i < testInputs.Length; i++)
{
double[] outputValues = network.Compute(testInputs[i]);
if (Compare(outputValues, testOutputs[i]))
{
correct++;
}
}
network.Save("deeplearning-countbits.net");
Console.WriteLine("Correct " + correct + "/" + testInputs.Length + ", " + Math.Round(((double)correct / (double)testInputs.Length * 100), 2) + "%");
}
static double Neural_Network(bool show)
{
double error = new double();
DataTable entireData = DataController.MakeDataTable("../../drug_consumption.txt");
Codification codebook = new Codification(entireData);
//"Alcohol", "Amfet", !!"Amyl", "Benzos", "Cofeine", "Cannabis", "Chocolate", "Coke", (1)"Crac", ///"Ecstasy", !!"Heroine",
// !!"Ketamine", //"LegalH", "LSD", !!"Meth", //"Mushrooms", "Nicotine", lol "Semeron", "VSA"
string LookingFor = "Heroine";
int good = 0;
string[][] outputs;
string[][] inputs = DataController.MakeString("../../drug_consumption_500.txt", out outputs);
string[][] testOutputs;
string[][] testInputs = DataController.MakeString("../../drug_consumption_500.txt", out testOutputs);
DataTable outputs1 = DataController.MakeDataFromString(outputs, "output");
DataTable inputs1 = DataController.MakeDataFromString(inputs, "input");
DataTable testOutputs1 = DataController.MakeDataFromString(testOutputs, "output");
DataTable testInputs1 = DataController.MakeDataFromString(testInputs, "input");
DataTable Isymbols = codebook.Apply(inputs1);
DataTable Osymbols = codebook.Apply(outputs1);
DataTable TIsymbols = codebook.Apply(testInputs1);
DataTable TOsymbols = codebook.Apply(testOutputs1);
double[][] inputsD = Isymbols.ToJagged <double>("Age", "Gender", "Education", "Country", "Eticnity", "Nscore", "Escore", "Oscore", "Ascore", "Cscore", "Impulsive", "SS");
double[][] outputsD = Osymbols.ToJagged <double>(LookingFor);
outputsD = DataController.convertDT(outputsD);
double[][] inputsT = TIsymbols.ToJagged <double>("Age", "Gender", "Education", "Country", "Eticnity", "Nscore", "Escore", "Oscore", "Ascore", "Cscore", "Impulsive", "SS");
double[][] outputsT = TOsymbols.ToJagged <double>(LookingFor);
outputsT = DataController.convertDT(outputsT);
DeepBeliefNetwork network = new DeepBeliefNetwork(inputs.First().Length, 10, 7);
new GaussianWeights(network, 0.1).Randomize();
network.UpdateVisibleWeights();
DeepBeliefNetworkLearning FirstLearner = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
int batchCount = Math.Max(1, inputs.Length / 100);
int[] groupsNew = Accord.Statistics.Classes.Random(inputsD.Length, batchCount);
double[][][] batchesNew = Accord.Statistics.Classes.Separate(inputsD, groupsNew);
double[][][] layerData;
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
FirstLearner.LayerIndex = layerIndex;
layerData = FirstLearner.GetLayerInput(batchesNew);
for (int i = 0; i < 500; i++)
{
error = FirstLearner.RunEpoch(layerData) / inputsD.Length;
if (i % 10 == 0 && show == true)
{
Console.WriteLine("Error value(" + LookingFor + ", test: " + i + ") = " + error);
}
}
}
var SecondLearner = new BackPropagationLearning(network)
{
LearningRate = 0.15,
Momentum = 0.7
};
EvolutionaryLearning teacher = new EvolutionaryLearning(network, 100);
for (int i = 0; i < 800; i++)
{
error = teacher.RunEpoch(inputsD, outputsD) / inputsD.Length;
if (i % 50 == 0 && show == true)
{
Console.WriteLine("Error value(" + LookingFor + ", test: " + i + ") = " + error);
}
}
for (int i = 0; i < 800; i++)
{
error = SecondLearner.RunEpoch(inputsD, outputsD) / inputsD.Length;
if (i % 10 == 0 && show == true)
{
Console.WriteLine("Error value(" + LookingFor + ", test: " + i + ") = " + error);
}
}
for (int i = 0; i < inputsD.Length; i++)
{
double[] outputValues = network.Compute(inputsT[i]);
if (outputValues.ToList().IndexOf(outputValues.Max()) == outputsT[i].ToList().IndexOf(outputsT[i].Max()))
{
good++;
}
}
if (show == true)
{
Console.WriteLine("Poprawność - " + Math.Round(((double)good / (double)inputsD.Length * 100), 4) + "%");
Console.ReadKey();
}
return(error);
}
//開始學習
public bool Run()
{
bool IsDone = false;
try
{
FlowDatas db = new FlowDatas();
(double[][] Inputs, double[][] Outputs)
= DeepLearningTools.FlowSampleToLearningData(db.FlowSampleStatistics.Where(c => c.BehaviorNumber != 0).ToArray());
db.Dispose();
//產生DBN網路
DBNetwork = new DeepBeliefNetwork(Inputs.First().Length,
(int)((Inputs.First().Length + Outputs.First().Length) / 1.5),
(int)((Inputs.First().Length + Outputs.First().Length) / 2),
Outputs.First().Length);
//亂數打亂整個網路參數
new GaussianWeights(DBNetwork, 0.1).Randomize();
DBNetwork.UpdateVisibleWeights();
//設定無監督學習組態
DeepBeliefNetworkLearning teacher
= new DeepBeliefNetworkLearning(DBNetwork)
{
Algorithm = (h, v, i) =>
new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.01,
Momentum = 0.5,
Decay = 0.001,
}
};
//設置批量輸入學習。
int batchCount1 = Math.Max(1, Inputs.Length / 10);
//創建小批量加速學習。
int[] groups1
= Accord.Statistics.Classes.Random(Inputs.Length, batchCount1);
double[][][] batches = Inputs.Subgroups(groups1);
//學習指定圖層的數據。
double[][][] layerData;
//運行無監督學習。
for (int layerIndex = 0; layerIndex < DBNetwork.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < 200; i++)
{
double error = teacher.RunEpoch(layerData) / Inputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
}
//對整個網絡進行監督學習，提供輸出分類。
var teacher2 = new ParallelResilientBackpropagationLearning(DBNetwork);
double error1 = double.MaxValue;
//運行監督學習。
for (int i = 0; i < 500; i++)
{
error1 = teacher2.RunEpoch(Inputs, Outputs) / Inputs.Length;
Console.WriteLine(i + ", Error = " + error1);
DBNetwork.Save(Path);
Console.WriteLine("Save Done");
}
DBNetwork.Save(Path);
Console.WriteLine("Save Done");
IsDone = true;
}
catch (Exception ex)
{
Debug.Write(ex.ToString());
}
return(IsDone);
}
static void Main(string[] args)
{
#if Cluster
// output file
List <string> outputLines = new List <string>();
DateTime timeStart = new DateTime();
// Some example documents.
string[] documents = new GetTweets().GetTweetsFromExcelFile("Train_NN.xlsx");
// Apply TF*IDF to the documents and get the resulting vectors.
double[][] inputs = TFIDF.Transform(documents, 0);
Console.WriteLine("time to transformation " + (DateTime.Now - timeStart));
outputLines.Add("time to transformation " + (DateTime.Now - timeStart));
Console.WriteLine("TFIDF transformation done...");
inputs = TFIDF.Normalize(inputs);
Console.WriteLine("time to Normalization " + (DateTime.Now - timeStart));
outputLines.Add("time to Normalization " + (DateTime.Now - timeStart));
Console.WriteLine("TFIDF Normalization done...");
//inputs = Accord.Math.Norm.Norm2(inputs);
string[] topics = TFIDF.Topics(documents, 5);
Console.WriteLine("time to topics " + (DateTime.Now - timeStart));
outputLines.Add("time to topics " + (DateTime.Now - timeStart));
Console.WriteLine("Topics gathered...");
//Random random = new Random();
//double[][] rand = new double[inputs.Length][];
//for (int i = 0; i < inputs.Length; i++)
//{
// rand[i] = new double[inputs[i].Length];
// for (int j = 0; j < inputs[i].Length; j++)
// {
// rand[i][j] = random.NextDouble();
// }
//}
//Console.WriteLine("time to generate random numbers " + (DateTime.Now - timeStart));
//outputLines.Add("time to topics " + (DateTime.Now - timeStart));
//Console.WriteLine("Randoms generated...");
KMeans cluster = new KMeans(topics.Length, Distance.Cosine);
//cluster.MaxIterations = 1;
//cluster.Randomize(rand);
int[] index = cluster.Compute(inputs);
Console.WriteLine("time to cluster " + (DateTime.Now - timeStart));
outputLines.Add("time to cluster " + (DateTime.Now - timeStart));
Console.WriteLine("Clustering done...");
//Accord.Statistics.Analysis.PrincipalComponentAnalysis pca = new Accord.Statistics.Analysis.PrincipalComponentAnalysis(inputs, Accord.Statistics.Analysis.AnalysisMethod.Center);
//pca.Compute();
//double[][] newinput = pca.Transform(inputs, 2);
//ScatterplotBox.Show("KMeans Clustering of Tweets", newinput, index).Hold();
for (double i = 0; i <= topics.Length; i++)
{
outputLines.Add(Convert.ToString(i + 1));
List <string> topicDecider = new List <string>();
string[] topicString;
int j = 0;
foreach (int x in index)
{
if (x == i + 1)
{
topicDecider.Add(documents[j]);
}
j++;
}
topicString = TFIDF.Topics(topicDecider.ToArray(), topicDecider.Count / 2);
if (topicString.Length == 0)
{
outputLines.Add("--------------------------------------------------------");
outputLines.Add("TOPIC: other");
outputLines.Add("--------------------------------------------------------");
}
else
{
outputLines.Add("--------------------------------------------------------");
outputLines.Add("TOPIC: " + topicString[0]);
outputLines.Add("--------------------------------------------------------");
}
j = 0;
foreach (int x in index)
{
if (x == i + 1)
{
outputLines.Add("Tweet ID " + j + ":\t" + documents[j]);
}
j++;
}
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
}
System.IO.File.WriteAllLines(@"Train_NN_2.txt", outputLines.ToArray());
Console.WriteLine("Output is written...");
#else
// output file
List <string> outputLines = new List <string>();
DateTime timeStart = new DateTime();
// Some example documents.
string[] documents_Train = new GetTweets().GetTweetsFromExcelFile("Train_NN.xlsx");
double[][] Train_Labels = new GetTweets().GetLabelsFromExcelFile("Train_Labels.xlsx");
// Apply TF*IDF to the documents and get the resulting vectors.
double[][] inputs = TFIDF.Transform(documents_Train, 0);
Console.WriteLine("time to transformation " + (DateTime.Now - timeStart));
outputLines.Add("time to transformation " + (DateTime.Now - timeStart));
Console.WriteLine("TFIDF transformation done...");
inputs = TFIDF.Normalize(inputs);
Console.WriteLine("time to Normalization " + (DateTime.Now - timeStart));
outputLines.Add("time to Normalization " + (DateTime.Now - timeStart));
Console.WriteLine("TFIDF Normalization done...");
//double[][] inputs;
double[][] train_input = new double[140][];
double[][] outputs;
double[][] testInputs = new double[1000 - 140][];
double[][] testOutputs = new double[1000 - 140][];
for (int i = 0; i < 140; i++)
{
train_input[i] = new double[inputs[i].Length];
for (int j = 0; j < inputs[i].Length; j++)
{
train_input[i][j] = inputs[i][j];
}
}
for (int i = 0; i < 1000 - 140; i++)
{
testInputs[i] = new double[inputs[i].Length];
for (int j = 0; j < inputs[i].Length; j++)
{
testInputs[i][j] = inputs[i][j];
}
}
// The first 500 data rows will be for training. The rest will be for testing.
//testInputs = inputs.Skip(500).ToArray();
//testOutputs = outputs.Skip(500).ToArray();
//inputs = inputs.Take(500).ToArray();
//outputs = outputs.Take(500).ToArray();
// Setup the deep belief network and initialize with random weights.
DeepBeliefNetwork network = new DeepBeliefNetwork(train_input.First().Length, 7);
new GaussianWeights(network, 0.1).Randomize();
network.UpdateVisibleWeights();
// Setup the learning algorithm.
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Setup batches of input for learning.
int batchCount = Math.Max(1, train_input.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Tools.RandomGroups(train_input.Length, batchCount);
double[][][] batches = train_input.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
// Unsupervised learning on each hidden layer, except for the output layer.
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < 200; i++)
{
double error = teacher.RunEpoch(layerData) / train_input.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
}
// Supervised learning on entire network, to provide output classification.
var teacher2 = new BackPropagationLearning(network)
{
LearningRate = 0.1,
Momentum = 0.5
};
//Transpose
double[][] Train_Labels_T = new double[140][];
for (int i = 0; i < 140; i++)
{
Train_Labels_T[i] = new double[7];
for (int j = 0; j < 7; j++)
{
Train_Labels_T[i][j] = Train_Labels[j][i];
}
}
// Run supervised learning.
for (int i = 0; i < 500; i++)
{
double error = teacher2.RunEpoch(train_input, Train_Labels_T) / train_input.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
outputLines.Add("time to Training " + (DateTime.Now - timeStart));
// Test the resulting accuracy.
double[][] outputValues = new double[testInputs.Length][];
for (int i = 0; i < testInputs.Length; i++)
{
outputValues[i] = network.Compute(testInputs[i]);
}
outputLines.Add("time to Testing/clustering " + (DateTime.Now - timeStart));
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
List <string> class1 = new List <string>();
List <string> class2 = new List <string>();
List <string> class3 = new List <string>();
List <string> class4 = new List <string>();
List <string> class5 = new List <string>();
List <string> class6 = new List <string>();
List <string> class7 = new List <string>();
//creating output file
for (int i = 0; i < documents_Train.Length; i++)
{
if (i < 10 && i > -1)
{
if (i == 0)
{
class1.Add("-------------------------------");
class1.Add("TOPIC: WEATHER");
class1.Add("-------------------------------");
}
class1.Add("Training_Tweet:\t" + documents_Train[i]);
}
if (i < 20 && i > 9)
{
if (i == 10)
{
class2.Add("-------------------------------");
class2.Add("TOPIC: MUSIC");
class2.Add("-------------------------------");
}
class2.Add("Training_Tweet:\t" + documents_Train[i]);
}
if (i < 30 && i > 19)
{
if (i == 20)
{
class3.Add("-------------------------------");
class3.Add("TOPIC: ITALY");
class3.Add("-------------------------------");
}
class3.Add("Training_Tweet:\t" + documents_Train[i]);
}
if (i < 40 && i > 29)
{
if (i == 30)
{
class4.Add("-------------------------------");
class4.Add("TOPIC: FOOD");
class4.Add("-------------------------------");
}
class4.Add("Training_Tweet:\t" + documents_Train[i]);
}
if (i < 50 && i > 39)
{
if (i == 40)
{
class5.Add("-------------------------------");
class5.Add("TOPIC: FASHION");
class5.Add("-------------------------------");
}
class5.Add("Training_Tweet:\t" + documents_Train[i]);
}
if (i < 60 && i > 49)
{
if (i == 50)
{
class6.Add("-------------------------------");
class6.Add("TOPIC: FOOTBALL");
class6.Add("-------------------------------");
}
class6.Add("Training_Tweet:\t" + documents_Train[i]);
}
if (i < 140 && i > 59)
{
if (i == 60)
{
class7.Add("-------------------------------");
class7.Add("TOPIC: OTHER");
class7.Add("-------------------------------");
}
class7.Add("Training_Tweet:\t" + documents_Train[i]);
}
if (i >= 140)
{
int what;
what = outputValues[i - 140].IndexOf(outputValues[i - 140].Max());
switch (what)
{
case 0:
class1.Add("Test_Tweet:\t" + documents_Train[i]);
break;
case 1:
class2.Add("Test_Tweet:\t" + documents_Train[i]);
break;
case 2:
class3.Add("Test_Tweet:\t" + documents_Train[i]);
break;
case 3:
class4.Add("Test_Tweet:\t" + documents_Train[i]);
break;
case 4:
class5.Add("Test_Tweet:\t" + documents_Train[i]);
break;
case 5:
class6.Add("Test_Tweet:\t" + documents_Train[i]);
break;
case 6:
class7.Add("Test_Tweet:\t" + documents_Train[i]);
break;
}
}
}
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.AddRange(class1);
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.AddRange(class2);
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.AddRange(class3);
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.AddRange(class4);
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.AddRange(class5);
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.AddRange(class6);
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
outputLines.AddRange(class7);
outputLines.Add("");
outputLines.Add("");
outputLines.Add("");
System.IO.File.WriteAllLines(@"Train_NN_With_Test_2.txt", outputLines.ToArray());
Console.Write("Press any key to quit ..");
#endif
Console.ReadKey();
}
private void DoTraining(string binaryFilePath, IOutputWriter writer)
{
double[][] inputs;
double[][] outputs;
// Load dataset.
inputs = DataManager.Load(binaryFilePath, out outputs);
// Setup the deep belief n etwork and initialize with random weights.
if (network == null)
{
network = DeepLearningNetworkBuilder.CreateAndGetNetwork(inputs.First().Length);
}
// Setup the learning algorithm.
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Setup batches of input for learning.
// int batchnumber = (int)totalCount / 10;
int batchCount = Math.Max(1, inputs.Length / 5);
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Classes.Random(inputs.Length, batchCount);
double[][][] batches = inputs.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
// Unsupervised learning on each hidden layer, except for the output layer.
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < (inputs.Length / 2); i++)
{
double error = teacher.RunEpoch(layerData) / inputs.Length;
writer?.WriteOutput(i + ", Error = " + error);
}
}
var teacher2 = new BackPropagationLearning(network)
{
LearningRate = 0.1,
Momentum = 0.5
};
// Run supervised learning.
for (int i = 0; i < inputs.Length; i++)
{
double error = teacher2.RunEpoch(inputs, outputs) / inputs.Length;
writer?.WriteOutput(i + ", Error = " + error);
}
}
public override double Train(Data info, float lr, float mom)
{
//Setup trainer using backpropagation.
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(DeepAccordNetwork)
{
Algorithm = (h, v, _) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = lr,
Momentum = mom,
Decay = 0.001,
}
};
// Setup batches of input for learning.
int batchCount = Math.Max(1, info.InputData.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Classes.Random(info.InputData.Length, batchCount); //Tools.RandomGroups(info.InputData.Length, batchCount);
double[][][] batches = Accord.Statistics.Classes.Separate(info.InputData, groups); //info.InputData.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
double error = double.PositiveInfinity;
double lastError;
// Unsupervised learning on each hidden layer, except for the output.
for (int layerIndex = 0; layerIndex < DeepAccordNetwork.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
do
{
lastError = error;
int i = 0;
while (i < 100)
{
error = teacher.RunEpoch(layerData);
i++;
}
} while (lastError - error > 0.0000001);
}
//Setup trainer using backpropagation.
BackPropagationLearning teacher2 = new BackPropagationLearning(DeepAccordNetwork)
{
LearningRate = lr,
Momentum = mom
};
do
{
lastError = error;
int i = 0;
while (i < 1000)
{
error = teacher2.RunEpoch(info.InputData, info.OutputData);
i++;
}
} while (lastError - error > 0.0000001);
return(error);
}
private void train_Click(object sender, EventArgs e)
{
double[][] inputs;
double[][] outputs;
double[][] testInputs;
double[][] testOutputs;
GetData(out inputs, out outputs, out testInputs, out testOutputs);
Stopwatch sw = Stopwatch.StartNew();
// Setup the deep belief network and initialize with random weights.
_network = new DeepBeliefNetwork(inputs.First().Length, LAYERS);
new GaussianWeights(_network, 0.1).Randomize();
_network.UpdateVisibleWeights();
// Setup the learning algorithm.
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(_network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Setup batches of input for learning.
int batchCount = Math.Max(1, inputs.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Tools.RandomGroups(inputs.Length, batchCount);
double[][][] batches = inputs.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
// Unsupervised learning on each hidden layer, except for the output layer.
for (int layerIndex = 0; layerIndex < _network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < UNSUPERVISED_EPOCHS; i++)
{
double error = teacher.RunEpoch(layerData) / inputs.Length;
if (i % 10 == 0)
{
label1.Text = "Layer: " + layerIndex + " Epoch: " + i + ", Error: " + error;
label1.Refresh();
}
}
}
// Supervised learning on entire network, to provide output classification.
var teacher2 = new BackPropagationLearning(_network)
{
LearningRate = 0.1,
Momentum = 0.5
};
// Run supervised learning.
for (int i = 0; i < SUPERVISED_EPOCHS; i++)
{
double error = teacher2.RunEpoch(inputs, outputs) / inputs.Length;
if (i % 10 == 0)
{
label1.Text = "Supervised: " + i + ", Error = " + error;
label1.Refresh();
}
}
// Test the resulting accuracy.
label1.Text = "";
int correct = 0;
for (int i = 0; i < testInputs.Length; i++)
{
double[] outputValues = _network.Compute(testInputs[i]);
int y = GetResult(outputValues);
int t = GetResult(testOutputs[i]);
label1.Text += "predicted: " + y + " actual: " + t + "\n";
label1.Refresh();
if (y == t)
{
correct++;
}
}
sw.Stop();
label1.Text = "Correct " + correct + "/" + testInputs.Length + ", " + Math.Round(((double)correct / (double)testInputs.Length * 100), 2) + "%";
label1.Text += "\nElapsed train+test time: " + sw.Elapsed;
label1.Refresh();
}
private static void Learn(string networkFile, double[][] inputs, double[][] outputs, double trainRate = 0.8)
{
var count = inputs.Length;
var n = (int)(count * trainRate);
var trainedInputs = inputs.Take(n).ToArray();
var trainedOutputs = outputs.Take(n).ToArray();
var testInputs = inputs.Skip(n).ToArray();
var testOutputs = outputs.Skip(n).ToArray();
Console.WriteLine($"trained items: {trainedInputs.Length}, tested items: {testInputs.Length}");
var network = new DeepBeliefNetwork(trainedInputs.First().Length, 10, trainedOutputs.First().Length);
new GaussianWeights(network, 0.1).Randomize();
network.UpdateVisibleWeights();
// Setup the learning algorithm.
var teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Setup batches of input for learning.
int batchCount = Math.Max(1, trainedInputs.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Classes.Random(trainedInputs.Length, batchCount);
double[][][] batches = trainedInputs.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
// Unsupervised learning on each hidden layer, except for the output layer.
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < 200; i++)
{
double error = teacher.RunEpoch(layerData) / trainedInputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
}
// Supervised learning on entire network, to provide output classification.
var teacher2 = new BackPropagationLearning(network)
{
LearningRate = 0.1,
Momentum = 0.5
};
// Run supervised learning.
for (int i = 0; i < Math.Min(2000, n); i++)
{
double error = teacher2.RunEpoch(trainedInputs, trainedOutputs) / trainedInputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
network.Save(networkFile);
Console.WriteLine($"save network: {networkFile}");
// Test the resulting accuracy.
Test(networkFile, testInputs, testOutputs);
}
public static void test()
{
//double[][] inputs;
//double[][] outputs;
//double[][] testInputs;
//double[][] testOutputs;
//// Load ascii digits dataset.
//inputs = DataManager.Load(@"../../../data/data.txt", out outputs);
//// The first 500 data rows will be for training. The rest will be for testing.
//testInputs = inputs.Skip(500).ToArray();
//testOutputs = outputs.Skip(500).ToArray();
//inputs = inputs.Take(500).ToArray();
//outputs = outputs.Take(500).ToArray();
//double[][] inputs = new double[4][] {
// new double[] {0, 0}, new double[] {0, 1},
// new double[] {1, 0}, new double[] {1, 1}
//};
//double[][] outputs = new double[4][] {
// new double[] {1, 0}, new double[] {0, 1},
// new double[] {0, 1}, new double[] {1, 0}
//};
double[][] inputs =
{
// input output
new double[] { 0, 1, 1, 0 }, // 0
new double[] { 0, 1, 0, 0 }, // 0
new double[] { 0, 0, 1, 0 }, // 0
new double[] { 0, 1, 1, 0 }, // 0
new double[] { 0, 1, 0, 0 }, // 0
new double[] { 1, 0, 0, 0 }, // 1
new double[] { 1, 0, 0, 0 }, // 1
new double[] { 1, 0, 0, 1 }, // 1
new double[] { 0, 0, 0, 1 }, // 1
new double[] { 0, 0, 0, 1 }, // 1
new double[] { 1, 1, 1, 1 }, // 2
new double[] { 1, 0, 1, 1 }, // 2
new double[] { 1, 1, 0, 1 }, // 2
new double[] { 0, 1, 1, 1 }, // 2
new double[] { 1, 1, 1, 1 }, // 2
};
double[][] outputs = // those are the class labels
{
new double[] { 1, 0, 0 },
new double[] { 1, 0, 0 },
new double[] { 1, 0, 0 },
new double[] { 1, 0, 0 },
new double[] { 1, 0, 0 },
new double[] { 0, 1, 0 },
new double[] { 0, 1, 0 },
new double[] { 0, 1, 0 },
new double[] { 0, 1, 0 },
new double[] { 0, 1, 0 },
new double[] { 0, 0, 1 },
new double[] { 0, 0, 1 },
new double[] { 0, 0, 1 },
new double[] { 0, 0, 1 },
new double[] { 0, 0, 1 },
};
// Setup the deep belief network and initialize with random weights.
Console.WriteLine(inputs.First().Length);
DeepBeliefNetwork network = new DeepBeliefNetwork(inputs.First().Length, 2, outputs.First().Length);
new GaussianWeights(network, 0.1).Randomize();
network.UpdateVisibleWeights();
// Setup the learning algorithm.
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Setup batches of input for learning.
int batchCount = Math.Max(1, inputs.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Tools.RandomGroups(inputs.Length, batchCount);
double[][][] batches = inputs.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
// Unsupervised learning on each hidden layer, except for the output layer.
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < 50000; i++)
{
double error = teacher.RunEpoch(layerData) / inputs.Length;
//if (i % 10 == 0)
//{
// Console.WriteLine(i + ", Error = " + error);
//}
}
}
// Supervised learning on entire network, to provide output classification.
var teacher2 = new Accord.Neuro.Learning.BackPropagationLearning(network)
{
LearningRate = 0.1,
Momentum = 0.5
};
// Run supervised learning.
for (int i = 0; i < 50000; i++)
{
double error = teacher2.RunEpoch(inputs, outputs) / inputs.Length;
//if (i % 10 == 0)
//{
// Console.WriteLine(i + ", Error = " + error);
//}
}
// Test the resulting accuracy.
//int correct = 0;
//for (int i = 0; i < inputs.Length; i++)
//{
// double[] outputValues = network.Compute(testInputs[i]);
// if (DataManager.FormatOutputResult(outputValues) == DataManager.FormatOutputResult(testOutputs[i]))
// {
// correct++;
// }
//}
//Console.WriteLine("Correct " + correct + "/" + inputs.Length + ", " + Math.Round(((double)correct / (double)inputs.Length * 100), 2) + "%");
//double[] probs = network.GenerateOutput(new double[] { 0, 0 });
//foreach (double p in probs)
//{
// Console.Write(p + ", ");
//}
for (int i = 0; i < inputs.Length; i++)
{
double[] output = network.Compute(inputs[i]);
int imax; output.Max(out imax);
Console.Write(imax + " -- ");
foreach (double p in output)
{
Console.Write(p + ", ");
}
Console.WriteLine("\n------------------");
}
}
public static void Excute2()
{
double[][] inputs;
double[][] outputs;
double[][] testInputs;
double[][] testOutputs;
// Load ascii digits dataset.
inputs = DataManager.Load(@"data.txt", out outputs);
// The first 500 data rows will be for training. The rest will be for testing. 第一个500数据用来训练，剩下的用来测试
testInputs = inputs.Skip(500).ToArray();
testOutputs = outputs.Skip(500).ToArray();
inputs = inputs.Take(500).ToArray();
outputs = outputs.Take(500).ToArray();
// Setup the deep belief network and initialize with random weights. 设置深度神经网络和初始化随机砝码
DeepBeliefNetwork network = new DeepBeliefNetwork(inputs.First().Length, 10, 10); //网络的输入数量Length 每个层中隐藏的神经元的数量10, 10
new GaussianWeights(network, 0.1).Randomize(); //高斯砝码 使用标准偏差。一般值在0.001—0.1范围内。 默认值为0.1。
//Randomize 使用高斯分布的网络的权重
network.UpdateVisibleWeights(); //通过复制隐藏层中权重的反向来更新可见层的权重。
// Setup the learning algorithm. 设置学习法则
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(network);
//自定义神经网络法则
// 设置用于指定和创建深度网络的每个层的学习算法的配置函数。Algorithm
teacher.Algorithm = (h, v, i) => {
return(new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1, //学习速率
Momentum = 0.5, //动力
Decay = 0.001, //腐烂
});
};
//teacher.Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
//{
// LearningRate = 0.1,
// Momentum = 0.5,
// Decay = 0.001,
//};
// Setup batches of input for learning.
int batchCount = System.Math.Max(1, inputs.Length / 100);//设置学习次数
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Classes.Random(inputs.Length, batchCount); //创建小批量 速度学习
double[][][] batches = inputs.Separate(groups); //分离
// Learning data for the specified layer.
double[][][] layerData; //为指定层 学习数据
// Unsupervised learning on each hidden layer, except for the output layer.除了输出层之外，在每个隐藏层上进行无监督学习。
//network.Machines.Count 在这个深网络的每一层上得到受限制的玻尔兹曼机器。
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
/*
* 获取训练数据所需的学习数据。
* 这个函数的返回应该被传递给No.Posial.SurvivyFieldWorksPr.RunEpoch（System，Pouth[2][]）。
* 去实践一个学习时代。
*/
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < 200; i++)//200次学习
{
var learningResult = teacher.RunEpoch(layerData);
double error = learningResult / inputs.Length;//RunEpoch运行纪元 Returns sum of learning errors.
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
}
// Supervised learning on entire network, to provide output classification.对整个网络进行监督学习，提供输出分类。
var teacher2 = new Neuro.Learning.BackPropagationLearning(network)
{
LearningRate = 0.1, //学习速率
Momentum = 0.5 //动力
};
// Run supervised learning.运行监督学习。
for (int i = 0; i < 500; i++)//500次学习
{
double error = teacher2.RunEpoch(inputs, outputs) / inputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
// Test the resulting accuracy. 测试结果的准确性
int correct = 0;
for (int i = 0; i < inputs.Length; i++)
{
var cp = testInputs[i].ToList();
var cpn = testInputs[i + 1];
foreach (var item in cpn)
{
cp.Add(item);
}
double[] outputValues = network.Compute(cp.ToArray());
//double[] outputValues = network.Compute(testInputs[i]);
if (DataManager.FormatOutputResult(outputValues) == DataManager.FormatOutputResult(testOutputs[i]))
{
correct++;
}
}
Console.WriteLine("Correct " + correct + "/" + inputs.Length + ", " + System.Math.Round(((double)correct / (double)inputs.Length * 100), 2) + "%");
Console.Write("Press any key to quit ..");
Console.ReadKey();
}
private void LearnForUpdatedModel(double[][] inputs, double[] outputDouble)
{
_modelMutex.WaitOne();
//updatingNetwork = network.DeepClone();
_modelMutex.ReleaseMutex();
//CreateTeachers(network);
var sampleSize = outputDouble.Length;
var inputToUse = inputs;
var outputToUse = outputDouble;
var outputData = new double[outputToUse.Length][];
for (var i = 0; i < outputToUse.Length; i++)
{
outputData[i] = new[] { outputToUse[i] };
}
LearningStopwatch.Reset();
LearningStopwatch.Start();
// Setup batches of input for learning.
var batchCount = Math.Max(1, inputToUse.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Tools.RandomGroups(inputToUse.Length, batchCount);
double[][][] batches = inputToUse.Subgroups(groups);
var epochsToUse = Epochs;
if (declineEpochs && sampleSize > declineRate)
{
epochsToUse = Math.Max(10, epochsToUse + declineRate - sampleSize);
}
// Unsupervised learning on each hidden layer, except for the output layer.
for (var layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
unsupervisedTeacher.LayerIndex = layerIndex;
var layerData = unsupervisedTeacher.GetLayerInput(batches);
for (var i = 0; i < epochsToUse / 3; i++)
{
unsupervisedTeacher.RunEpoch(layerData);
}
}
// Run supervised learning.
for (var i = 0; i < epochsToUse; i++)
{
supervisedTeacher.RunEpoch(inputToUse, outputData);
}
//updatingNetwork.UpdateVisibleWeights();
LearningStopwatch.Stop();
LastLearnTime = LearningStopwatch.ElapsedMilliseconds;
LearningStopwatch.Reset();
//return updatingNetwork;
}
public void ExampleTest1()
{
Accord.Math.Tools.SetupGenerator(0);
// We'll use a simple XOR function as input.
double[][] inputs =
{
new double[] { 0, 0 }, // 0 xor 0
new double[] { 0, 1 }, // 0 xor 1
new double[] { 1, 0 }, // 1 xor 0
new double[] { 1, 1 }, // 1 xor 1
};
// XOR output, corresponding to the input.
double[][] outputs =
{
new double[] { 0 }, // 0 xor 0 = 0
new double[] { 1 }, // 0 xor 1 = 1
new double[] { 1 }, // 1 xor 0 = 1
new double[] { 0 }, // 1 xor 1 = 0
};
// Setup the deep belief network (2 inputs, 3 hidden, 1 output)
DeepBeliefNetwork network = new DeepBeliefNetwork(2, 3, 1);
// Initialize the network with Gaussian weights
new GaussianWeights(network, 0.1).Randomize();
// Update the visible layer with the new weights
network.UpdateVisibleWeights();
// Setup the learning algorithm.
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Unsupervised learning on each hidden layer, except for the output.
for (int i = 0; i < network.Layers.Length - 1; i++)
{
teacher.LayerIndex = i;
// Compute the learning data with should be used
var layerInput = teacher.GetLayerInput(inputs);
// Train the layer iteratively
for (int j = 0; j < 5000; j++)
{
teacher.RunEpoch(layerInput);
}
}
// Supervised learning on entire network, to provide output classification.
var backpropagation = new BackPropagationLearning(network)
{
LearningRate = 0.1,
Momentum = 0.5
};
// Run supervised learning.
for (int i = 0; i < 5000; i++)
{
backpropagation.RunEpoch(inputs, outputs);
}
// Test the resulting accuracy.
int correct = 0;
for (int i = 0; i < inputs.Length; i++)
{
double[] outputValues = network.Compute(inputs[i]);
double outputResult = outputValues.First() >= 0.5 ? 1 : 0;
if (outputResult == outputs[i].First())
{
correct++;
}
}
Assert.AreEqual(4, correct);
}
static void Main(string[] args)
{
double[][] inputs;
double[][] outputs;
double[][] testInputs;
double[][] testOutputs;
const int SampleTrainingCount = 120;
const int SampleTestCount = 30;
// Load ascii digits dataset.
inputs = DataManager.LoadCSV(@"../../../data/iris.data", out outputs);
//inputs = DataManager.Load(@"../../../data/data.txt", out outputs);
// The first SampleTrainingCount data rows will be for training. The rest will be for testing.
testInputs = inputs.Skip(SampleTrainingCount).ToArray();
testOutputs = outputs.Skip(SampleTrainingCount).ToArray();
inputs = inputs.Take(SampleTrainingCount).ToArray();
outputs = outputs.Take(SampleTrainingCount).ToArray();
// Setup the deep belief network and initialize with random weights.
DeepBeliefNetwork network = new DeepBeliefNetwork(inputs.First().Length, 10, 1);
new GaussianWeights(network, 0.1).Randomize();
network.UpdateVisibleWeights();
// Setup the learning algorithm.
DeepBeliefNetworkLearning teacher = new DeepBeliefNetworkLearning(network)
{
Algorithm = (h, v, i) => new ContrastiveDivergenceLearning(h, v)
{
LearningRate = 0.1,
Momentum = 0.5,
Decay = 0.001,
}
};
// Setup batches of input for learning.
int batchCount = Math.Max(1, inputs.Length / 100);
// Create mini-batches to speed learning.
int[] groups = Accord.Statistics.Tools.RandomGroups(inputs.Length, batchCount);
double[][][] batches = inputs.Subgroups(groups);
// Learning data for the specified layer.
double[][][] layerData;
// Unsupervised learning on each hidden layer, except for the output layer.
for (int layerIndex = 0; layerIndex < network.Machines.Count - 1; layerIndex++)
{
teacher.LayerIndex = layerIndex;
layerData = teacher.GetLayerInput(batches);
for (int i = 0; i < 200; i++)
{
double error = teacher.RunEpoch(layerData) / inputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
}
// Supervised learning on entire network, to provide output classification.
var teacher2 = new BackPropagationLearning(network)
{
LearningRate = 0.1,
Momentum = 0.5
};
// Run supervised learning.
for (int i = 0; i < SampleTrainingCount; i++)
{
double error = teacher2.RunEpoch(inputs, outputs) / inputs.Length;
if (i % 10 == 0)
{
Console.WriteLine(i + ", Error = " + error);
}
}
// Test the resulting accuracy. SampleTestCount item
int correct = 0;
for (int i = 0; i < SampleTestCount; i++)
{
double[] outputValues = network.Compute(testInputs[i]);
if (DataManager.FormatOutputResult(outputValues) == DataManager.FormatOutputResult(testOutputs[i]))
{
correct++;
}
}
Console.WriteLine("Correct " + correct + "/" + SampleTestCount + ", " + Math.Round(((double)correct / (double)SampleTestCount * 100), 2) + "%");
Console.Write("Press any key to quit ..");
Console.ReadKey();
}