/// <summary>
/// Evaluate memory.
/// </summary>
private void EvalMemory()
{
BasicNeuralDataSet training = RandomTrainingFactory.Generate(
1000, 10000, 10, 10, -1, 1);
long stop = (10 * Evaluate.MILIS);
int record = 0;
INeuralDataPair pair = BasicNeuralDataPair.CreatePair(10, 10);
int iterations = 0;
Stopwatch watch = new Stopwatch();
watch.Start();
while (watch.ElapsedMilliseconds < stop)
{
iterations++;
training.GetRecord(record++, pair);
if (record >= training.Count)
{
record = 0;
}
}
iterations /= 100000;
this.report.Report(EncogBenchmark.STEPS, EncogBenchmark.STEP3,
"Memory dataset, result: " + Format.FormatInteger(iterations));
this.memoryScore = iterations;
}
public void CalculateAnswer()
{
var values = new List <double[]>();
var answers = new List <double[]>();
var outputSize = parameters.ProblemType == ProblemTypeEnum.Classification ? classCount : parameters.CountOutput;
foreach (var input in testSet)
{
var output = new double[outputSize];
network.Compute(input, output);
values.Add(input);
answers.Add(output);
}
for (int i = 0; i < values.Count; i++)
{
for (int argc = 0; argc < values[i].Length; argc++)
{
values[i][argc] = ArgNormalize[argc].DeNormalize(values[i][argc]);
}
}
if (parameters.ProblemType == ProblemTypeEnum.Regression)
{
for (int i = 0; i < values.Count; i++)
{
answers[i][0] = YNormalize.DeNormalize(answers[i][0]);
}
// DenormalizationRegg(values, false);
//DenormalizationRegg(answers,true);
}
AnswerSet = new BasicNeuralDataSet(values.ToArray(), answers.ToArray());
}
static void Main(string[] args)
{
int dimension = 2;
int numNeuronsPerDimension = 4;
double volumeNeuronWidth = 2.0 / numNeuronsPerDimension;;
bool includeEdgeRBFs = true;
RBFNetwork n = new RBFNetwork(dimension, numNeuronsPerDimension, 1, RBFEnum.Gaussian);
n.SetRBFCentersAndWidthsEqualSpacing(0, 1, RBFEnum.Gaussian, volumeNeuronWidth, includeEdgeRBFs);
INeuralDataSet trainingSet = new BasicNeuralDataSet(XORInput, XORIdeal);
SVDTraining train = new SVDTraining(n, trainingSet);
int epoch = 1;
do
{
train.Iteration();
Console.WriteLine(@"Epoch #" + epoch + @" Error:" + train.Error);
epoch++;
} while ((epoch < 1) && (train.Error > 0.001));
Console.WriteLine(@"Neural Network Results:");
foreach (IMLDataPair pair in trainingSet)
{
IMLData output = n.Compute(pair.Input);
Console.WriteLine(pair.Input[0] + @"," + pair.Input[1]
+ @", actual=" + output[0] + @",ideal=" + pair.Ideal[0]);
}
Console.Read();
}
public double MeasurePerformance(IMLRegression network, BasicNeuralDataSet dataset, IActivationFunction activationFunction)
{
int correctBits = 0;
const float Threshold = 0.0f;
if (!(activationFunction is ActivationTANH))
{
throw new ArgumentException("Bad activation function");
}
int n = (int)dataset.Count;
for (int inputIndex = 0; inputIndex < n; inputIndex++)
{
var actualOutputs = network.Compute(dataset.Data[inputIndex].Input);
for (int outputIndex = 0, k = actualOutputs.Count; outputIndex < k; outputIndex++)
{
if (IsBothPositiveBits(actualOutputs[outputIndex], dataset.Data[inputIndex].Ideal[outputIndex], Threshold) ||
IsBothNegativeBits(actualOutputs[outputIndex], dataset.Data[inputIndex].Ideal[outputIndex], Threshold))
{
correctBits++;
}
}
}
long totalBitsCount = dataset.Count * dataset.Data[0].Ideal.Count;
return((double)correctBits / totalBitsCount);
}
public static void Train(PointsConverted pointsConvertedA, PointsConverted pointsConvertedB = null, bool debug = true)
{
network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, N_input));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, N_hidden));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, N_output));
network.Structure.FinalizeStructure();
network.Reset();
pointsConvertedA.Juntar(pointsConvertedB);
var trainingSet = new BasicNeuralDataSet(pointsConvertedA.entrada, pointsConvertedA.saida);
var train = new Backpropagation(network, trainingSet);
var epoch = 0;
do
{
train.Iteration();
if (debug)
{
Console.WriteLine("Epoch #" + epoch + " Error:" + train.Error);
}
epoch++;
}while ((epoch <= 20000) || (train.Error > 0.001));
}
public static void Main()
{
BasicNetwork network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 2));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 6));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 1));
network.Structure.FinalizeStructure();
network.Reset();
INeuralDataSet trainingSet = new BasicNeuralDataSet(XorInput, XorIdeal);
ITrain train = new ResilientPropagation(network, trainingSet);
int epoch = 1;
var timer = Stopwatch.StartNew();
do
{
train.Iteration();
epoch++;
} while ((epoch < 50000) && (train.Error > 0.0001));
timer.Stop();
Console.WriteLine("Neural Network Results:");
foreach (var pair in trainingSet)
{
var output = network.Compute(pair.Input);
Console.WriteLine(pair.Input[0] + "," + pair.Input[1]
+ ", actual=" + output[0] + ", ideal=" + pair.Ideal[0]);
}
Console.WriteLine($"Completed {epoch} epochs in {timer.Elapsed} ({(float)timer.ElapsedMilliseconds / epoch} ms per epoch)");
Console.ReadLine();
}
private double CalculateClassificationError(BasicNeuralDataSet trainingSet)
{
int errorCount = 0;
foreach (var trainData in trainingSet)
{
IMLData output = _network.Compute(trainData.Input);
IMLData ideal = trainData.Ideal;
double maxValue = Double.MinValue;
int maxIndex = 0;
for (int i = 0; i < output.Count; ++i)
{
if (maxValue < output[i])
{
maxValue = output[i];
maxIndex = i;
}
}
if (Math.Abs(ideal[maxIndex] - 1) > 0.0001)
{
errorCount++;
}
}
return((double)errorCount / trainingSet.Count);
}
/// <summary>
/// Load the specified Encog object from an XML reader.
/// </summary>
/// <param name="xmlIn">The XML reader to use.</param>
/// <returns>The loaded object.</returns>
public IEncogPersistedObject Load(ReadXML xmlIn)
{
String name = xmlIn.LastTag.Attributes[
EncogPersistedCollection.ATTRIBUTE_NAME];
String description = xmlIn.LastTag.GetAttributeValue(
EncogPersistedCollection.ATTRIBUTE_DESCRIPTION);
this.currentDataSet = new BasicNeuralDataSet();
this.currentDataSet.Name = name;
this.currentDataSet.Description = description;
while (xmlIn.ReadToTag())
{
if (xmlIn.IsIt(BasicNeuralDataSetPersistor.TAG_ITEM, true))
{
HandleItem(xmlIn);
}
else if (xmlIn.IsIt(EncogPersistedCollection.TYPE_TRAINING, false))
{
break;
}
}
return(this.currentDataSet);
}
static void Main(string[] args)
{
INeuralDataSet trainingSet = new BasicNeuralDataSet(AndInput, AndIdeal);
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 2));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 2));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 1));
network.Structure.FinalizeStructure();
network.Reset();
ITrain train = new ResilientPropagation(network, trainingSet);
int epoch = 1;
do
{
train.Iteration();
Console.WriteLine($"Epoch no {epoch}. Error: {train.Error}");
epoch++;
} while ((epoch < MaxEpoch) && (train.Error > AcceptableError));
Console.WriteLine("\nAnd function Results:");
foreach (IMLDataPair pair in trainingSet)
{
IMLData output = network.Compute(pair.Input);
Console.WriteLine($"{pair.Input[0]} AND {pair.Input[1]} should be: {pair.Ideal[0]} actual value is: {output[0]}");
}
Console.ReadKey();
}
public void Prune()
{
INeuralDataSet trainingSet = new BasicNeuralDataSet(networkInput, networkIdealOutput);
FeedForwardPattern pattern = new FeedForwardPattern();
pattern.InputNeurons = INPUT_NEURONS;
pattern.OutputNeurons = OUTPUT_NEURONS;
if (ACTIVIATION_FUNCTION == 1)
pattern.ActivationFunction = new ActivationSigmoid();
else if (ACTIVIATION_FUNCTION == 2)
pattern.ActivationFunction = new ActivationTANH();
else
throw new System.Exception("Only 2 activation functions have been impletemented.");
PruneIncremental prune = new PruneIncremental(trainingSet, pattern, 200, new ConsoleStatusReportable());
prune.AddHiddenLayer(10, 40);
prune.AddHiddenLayer(0, 30);
prune.Process();
network = prune.BestNetwork;
Console.WriteLine("Prune process complete.");
}
static void Main(string[] args)
{
int dimension = 2; // XORInput provides two-dimensional inputs. Not 8.
/*
* If XORInput is 8 dimensional it should be like this:
*
* public static double[][] XORInput = {
* new[] {0.0, 0.0,0.0, 0.0,0.0, 0.0,0.0, 0.0},
* .
* .
* .*/
int numNeuronsPerDimension = 4; // could be also 16, 64, 256. I suppose it should accept 8, 32 but it needs additional investigation
double volumeNeuronWidth = 2.0 / numNeuronsPerDimension;
bool includeEdgeRBFs = true;
RBFNetwork n = new RBFNetwork(dimension, numNeuronsPerDimension, 1, RBFEnum.Gaussian);
n.SetRBFCentersAndWidthsEqualSpacing(0, 1, RBFEnum.Gaussian, volumeNeuronWidth, includeEdgeRBFs);
//n.RandomizeRBFCentersAndWidths(0, 1, RBFEnum.Gaussian);
INeuralDataSet trainingSet = new BasicNeuralDataSet(XORInput, XORIdeal);
SVDTraining train = new SVDTraining(n, trainingSet);
int epoch = 1;
do
{
train.Iteration();
Console.WriteLine("Epoch #" + epoch + " Error:" + train.Error);
epoch++;
} while ((epoch < 1) && (train.Error > 0.001));
}
/// <summary>
/// Generate a random training set.
/// </summary>
/// <param name="seed">The seed value to use, the same seed value will always produce
/// the same results.</param>
/// <param name="count">How many training items to generate.</param>
/// <param name="inputCount">How many input numbers.</param>
/// <param name="idealCount">How many ideal numbers.</param>
/// <param name="min">The minimum random number.</param>
/// <param name="max">The maximum random number.</param>
/// <returns>The random training set.</returns>
public static BasicNeuralDataSet Generate(long seed,
int count, int inputCount,
int idealCount, double min, double max)
{
LinearCongruentialGenerator rand =
new LinearCongruentialGenerator(seed);
BasicNeuralDataSet result = new BasicNeuralDataSet();
for (int i = 0; i < count; i++)
{
INeuralData inputData = new BasicNeuralData(inputCount);
for (int j = 0; j < inputCount; j++)
{
inputData.Data[j] = rand.Range(min, max);
}
INeuralData idealData = new BasicNeuralData(idealCount);
for (int j = 0; j < idealCount; j++)
{
idealData[j] = rand.Range(min, max);
}
BasicNeuralDataPair pair = new BasicNeuralDataPair(inputData,
idealData);
result.Add(pair);
}
return(result);
}
private void TrainNetwork(DateTime trainFrom, DateTime trainTo, TrainingStatus status)
{
if (_input == null || _ideal == null)
{
CreateTrainingSets(trainFrom, trainTo);
}
_trainThread = Thread.CurrentThread;
int epoch = 1;
ITrain train = null;
try
{
var trainSet = new BasicNeuralDataSet(_input, _ideal);
train = new ResilientPropagation(_network, trainSet);
double error;
do
{
train.Iteration();
error = train.Error;
status?.Invoke(epoch, error, TrainingAlgorithm.Resilient);
epoch++;
} while (error > MaxError);
}
catch (ThreadAbortException) { _trainThread = null; }
finally
{
train?.FinishTraining();
}
_trainThread = null;
}
public static BasicNetwork TrainNetwork(BasicNetwork network, double[][] trainingData, double[][] ideals,
int maxIterationsBeforeCompletion = 5000, double tolerance = 0.001, string connectionId = null,
Func <string, string, bool, bool> OutputData = null)
{
//Check we have data and a model to train
if (trainingData.Any() && ideals.Any() && network != null)
{
//Create Dataset - data and correct classifications (matched by position)
INeuralDataSet trainingSet = new BasicNeuralDataSet(trainingData, ideals);
//Proagate the data through the network
ITrain train = new ResilientPropagation(network, trainingSet);
//Set the iteration count to 0
var epoch = 0;
//Train
do
{
train.Iteration();
//If the delegate is defined, output the progress to it
if (OutputData != null)
{
OutputData(connectionId, "Epoch #" + epoch + " Error:" + train.Error, true);
}
epoch++;
} while ((epoch < maxIterationsBeforeCompletion) && (train.Error > tolerance));
}
//Return the trained network
return(network);
}
public void LoadCSVFileTrainingClass()
{
var path = parameters.TrainFile;
var csvRead = new ReadCSV(new FileStream(path, FileMode.Open), true, CSVFormat.DecimalPoint);
var values = new List <double[]>();
var classes = new List <double[]>();
while (csvRead.Next())
{
values.Add(new double[2] {
csvRead.GetDouble(0), csvRead.GetDouble(1)
});
classes.Add(new double[1] {
csvRead.GetDouble(2)
});
}
csvRead.Close();
var min = parameters.FunctionType == FunctionTypeEnum.Unipolar ? 0d : -1d;
ArgNormalize = new NormalizedField[2] {
new NormalizedField(NormalizationAction.Normalize, "X", values.Max(v => v[0]), values.Min(v => v[0]), 1.0, min),
new NormalizedField(NormalizationAction.Normalize, "Y", values.Max(v => v[1]), values.Min(v => v[1]), 1.0, min)
};
for (int i = 0; i < values.Count; i++)
{
for (int argc = 0; argc < values[i].Length; argc++)
{
values[i][argc] = ArgNormalize[argc].Normalize(values[i][argc]);
}
}
//Normalization(values);
classCount = classes.Select(c => c[0]).Distinct().Count();
var normalizeClasses = new List <double[]>();
for (int i = 0; i < classes.Count; ++i)
{
var newClasses = new double[classCount];
for (int j = 0; j < newClasses.Length; j++)
{
newClasses[j] = min;
}
newClasses[(int)classes[i][0] - 1] = 1;// dodoac normalizacje na -1
normalizeClasses.Add(newClasses);
}
var trainSetCount = (int)((double)values.Count * ((100.0 - 15) / 100));
values.Shuffle();
normalizeClasses.Shuffle();
MyExtensions.ResetStableShuffle();
TrainSet = new BasicNeuralDataSet(values.Take(trainSetCount).ToArray(),
normalizeClasses.Take(trainSetCount).ToArray());
ValidationSet = new BasicNeuralDataSet(values.Skip(trainSetCount).ToArray(),
normalizeClasses.Skip(trainSetCount).ToArray());
}
public Task <Network> Train(NetworkConfiguration networkConfiguration, int[] spectralImagesIndexesToConsider, TrainingCallback callback)
{
var network = networkFactory.Create(
networkConfiguration.ActivationFunction,
DefaultFingerprintSize,
networkConfiguration.HiddenLayerCount,
networkConfiguration.OutputCount);
var spectralImagesToTrain = trainingDataProvider.GetSpectralImagesToTrain(
spectralImagesIndexesToConsider, (int)System.Math.Pow(2, networkConfiguration.OutputCount));
var trainingSet = trainingDataProvider.MapSpectralImagesToBinaryOutputs(
spectralImagesToTrain, networkConfiguration.OutputCount);
normalizeStrategy.NormalizeInputInPlace(networkConfiguration.ActivationFunction, trainingSet.Inputs);
normalizeStrategy.NormalizeOutputInPlace(networkConfiguration.ActivationFunction, trainingSet.Outputs);
return(Task.Factory.StartNew(
() =>
{
var dataset = new BasicNeuralDataSet(trainingSet.Inputs, trainingSet.Outputs);
var learner = new Backpropagation(network, dataset);
double correctOutputs = 0.0;
for (int idynIndex = 0; idynIndex < Idyn; idynIndex++)
{
correctOutputs = networkPerformanceMeter.MeasurePerformance(
network, dataset, networkConfiguration.ActivationFunction);
callback(TrainingStatus.OutputReordering, correctOutputs, learner.Error, idynIndex * Edyn);
var bestPairs = GetBestPairsForReordering(
(int)System.Math.Pow(2, networkConfiguration.OutputCount), network, spectralImagesToTrain, trainingSet);
ReorderOutputsAccordingToBestPairs(bestPairs, trainingSet, dataset);
for (int edynIndex = 0; edynIndex < Edyn; edynIndex++)
{
correctOutputs = networkPerformanceMeter.MeasurePerformance(
network, dataset, networkConfiguration.ActivationFunction);
callback(
TrainingStatus.RunningDynamicEpoch,
correctOutputs,
learner.Error,
(idynIndex * Edyn) + edynIndex);
learner.Iteration();
}
}
for (int efixedIndex = 0; efixedIndex < Efixed; efixedIndex++)
{
correctOutputs = networkPerformanceMeter.MeasurePerformance(
network, dataset, networkConfiguration.ActivationFunction);
callback(
TrainingStatus.FixedTraining, correctOutputs, learner.Error, (Idyn * Edyn) + efixedIndex);
learner.Iteration();
}
network.ComputeMedianResponses(trainingSet.Inputs, TrainingSongSnippets);
callback(TrainingStatus.Finished, correctOutputs, learner.Error, (Idyn * Edyn) + Efixed);
return network;
}));
}
private static EncogTrainingResponse TrainNetwork2(BasicNetwork network, TrainingData training, double maxError, CancellationToken cancelToken, double?maxSeconds = null)
{
//TODO: When the final layer is softmax, the error seems to be higher. Probably because the training outputs need to be run through softmax
const int MAXITERATIONS = 5000;
INeuralDataSet trainingSet = new BasicNeuralDataSet(training.Input, training.Output);
ITrain train = new ResilientPropagation(network, trainingSet);
DateTime startTime = DateTime.UtcNow;
TimeSpan?maxTime = maxSeconds != null?TimeSpan.FromSeconds(maxSeconds.Value) : (TimeSpan?)null;
bool success = false;
//List<double> log = new List<double>();
int iteration = 1;
double error = double.MaxValue;
while (true)
{
if (cancelToken.IsCancellationRequested)
{
break;
}
train.Iteration();
error = train.Error;
//log.Add(error);
iteration++;
if (double.IsNaN(error))
{
break;
}
else if (error < maxError)
{
success = true;
break;
}
else if (iteration >= MAXITERATIONS)
{
break;
}
else if (maxTime != null && DateTime.UtcNow - startTime > maxTime)
{
break;
}
}
//string logExcel = string.Join("\r\n", log); // paste this into excel and chart it to see the error trend
train.FinishTraining();
return(new EncogTrainingResponse(network, success, error, iteration, (DateTime.UtcNow - startTime).TotalSeconds));
}
/// <summary>
/// Load the binary dataset to memory. Memory access is faster.
/// </summary>
/// <returns>A memory dataset.</returns>
public INeuralDataSet LoadToMemory()
{
BasicNeuralDataSet result = new BasicNeuralDataSet();
foreach (INeuralDataPair pair in this)
{
result.Add(pair);
}
return(result);
}
private BasicNeuralDataSet PrepareRegressionSet(double[][] data)
{
var normalizedData = _normalizer.Normalize(data);
var input = normalizedData.Select(row => row.Take(_inputLayerSize).ToArray()).ToArray();
var ideal = normalizedData.Select(row => row.Skip(_inputLayerSize).ToArray()).ToArray();
var trainingSet = new BasicNeuralDataSet(input, ideal);
return(trainingSet);
}
private BasicNetwork BuildModel(TimeSeries simulatedData, out double scale_factor)
{
double[][] inputs = new double[simulatedData.Count][];
scale_factor = 0;
for (int i = 0; i < simulatedData.Count; ++i)
{
if (scale_factor < simulatedData[i])
{
scale_factor = simulatedData[i];
}
}
for (int i = 0; i < simulatedData.Count; ++i)
{
inputs[i] = new double[mWindowSize];
for (int j = 0; j < mWindowSize; ++j)
{
int index = i - (mWindowSize - j);
if (index >= 0)
{
inputs[i][j] = simulatedData[index] / scale_factor;
}
}
}
double[][] outputs = new double[simulatedData.Count][];
for (int i = 0; i < simulatedData.Count; ++i)
{
outputs[i] = new double[1] {
simulatedData[i] / scale_factor
};
}
INeuralDataSet trainingSet = new BasicNeuralDataSet(inputs, outputs);
BasicNetwork network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, mWindowSize));
for (int i = 0; i < mNumberHiddenLayer; ++i)
{
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, mWindowSize));
}
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 1));
network.Structure.FinalizeStructure();
network.Reset();
Train(network, trainingSet);
return(network);
}
public override List <double> Train(List <Tuple <List <double>, List <double> > > trainSet, int iterations)
{
var errors = new List <double>(iterations);
INeuralDataSet trainingSet = new BasicNeuralDataSet(trainSet.Select(e => e.Item1.ToArray()).ToArray(), trainSet.Select(e => e.Item2.ToArray()).ToArray());
ITrain train = new Backpropagation(network, trainingSet, this.learningRate, this.momentum);
var epoch = 0;
do
{
train.Iteration();
epoch++;
errors.Add(train.Error);
} while (epoch < iterations);
return(errors);
}
public static void TestNetwork(BasicNetwork network, double[][] trainingData, double[][] ideals,
string connectionId, Func <string, string, bool, bool> OutputData = null)
{
//Create Dataset - data and correct classifications (matched by position)
var trainingSet = new BasicNeuralDataSet(trainingData, ideals);
foreach (var pair in trainingSet)
{
var output = network.Compute(pair.Input);
//Columns: |Rain|,|Wind|,|Sun|,|Cloud/Overcast|,|Snow|
if (OutputData != null)
{
OutputData(connectionId, FormatOutput(pair, output), false);
}
}
}
/// <summary>
/// Evaluate disk.
/// </summary>
private void EvalBinary()
{
String file = "temp.egb";
BasicNeuralDataSet training = RandomTrainingFactory.Generate(
1000, 10000, 10, 10, -1, 1);
// create the binary file
File.Delete(file);
BufferedNeuralDataSet training2 = new BufferedNeuralDataSet(file);
training2.Load(training);
long stop = (10 * Evaluate.MILIS);
int record = 0;
INeuralDataPair pair = BasicNeuralDataPair.CreatePair(10, 10);
Stopwatch watch = new Stopwatch();
watch.Start();
int iterations = 0;
while (watch.ElapsedMilliseconds < stop)
{
iterations++;
training2.GetRecord(record++, pair);
if (record >= training2.Count)
{
record = 0;
}
}
training2.Close();
iterations /= 100000;
this.report.Report(EncogBenchmark.STEPS, EncogBenchmark.STEP4,
"Disk(binary) dataset, result: "
+ Format.FormatInteger(iterations));
File.Delete(file);
this.binaryScore = iterations;
}
public void MainTestEncog()
{
int classesCount = 0;
int attributesCount = 0;
var epoch = 0;
var trainData = CsvParser.Parse("./../../../DataSets/data.train.csv", ref classesCount, ref attributesCount).NormalizedData;
var testData = CsvParser.Parse("./../../../DataSets/data.train.csv", ref classesCount, ref attributesCount).NormalizedData;
var correct = 0;
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 2));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 4));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 5));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 3));
network.Structure.FinalizeStructure();
network.Reset();
INeuralDataSet trainingSet = new BasicNeuralDataSet(trainData.Select(e => e.Item1.ToArray()).ToArray(), trainData.Select(e => e.Item2.ToArray()).ToArray());
ITrain train = new Backpropagation(network, trainingSet, 0.3, 0.6);
//ITrain train = new ResilientPropagation(network, trainingSet);
do
{
train.Iteration();
epoch++;
} while ((epoch < 15000) && (train.Error > 0.001));
foreach (IMLDataPair pair in trainingSet)
{
var output = network.Compute(pair.Input);
//pair.Ideal
if (Network.GetClass(new List <double>()
{
output[0], output[1], output[2]
}) == Network.GetClass(new List <double>()
{
pair.Ideal[0], pair.Ideal[1], pair.Ideal[2]
}))
{
correct++;
}
}
}
static void Main(string[] args)
{
INeuralDataSet trainingSet = new BasicNeuralDataSet(AndInput, AndIdeal);
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationRamp(), true, 25));
network.AddLayer(new BasicLayer(new ActivationRamp(), true, 75));
network.AddLayer(new BasicLayer(new ActivationRamp(), true, 50));
network.AddLayer(new BasicLayer(new ActivationRamp(), true, 20));
network.Structure.FinalizeStructure();
network.Reset();
ITrain train = new Backpropagation(network, trainingSet, 0.02, 0.3);
int epoch = 1;
do
{
train.Iteration();
Console.WriteLine($"{train.Error}");
epoch++;
} while ((epoch < MaxEpoch) && (train.Error > AcceptableError));
var input = new BasicMLData(new double[25] {
1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1
});
int best = network.Winner(input);
Console.WriteLine($"Rozpoznano: {_literki[best]}");
foreach (IMLDataPair pair in trainingSet)
{
IMLData output = network.Compute(pair.Input);
Console.WriteLine($"wynik : { output[0].ToString("0.000")} {output[1].ToString("0.000")} {output[2].ToString("0.000")} {output[3].ToString("0.000")} {output[4].ToString("0.000")} {output[5].ToString("0.000")} {output[6].ToString("0.000")} {output[7].ToString("0.000")} {output[8].ToString("0.000")} {output[9].ToString("0.000")} {output[10].ToString("0.000")} {output[11].ToString("0.000")} {output[12].ToString("0.000")} {output[13].ToString("0.000")} {output[14].ToString("0.000")} {output[15].ToString("0.000")} {output[16].ToString("0.000")} {output[17].ToString("0.000")} {output[18].ToString("0.000")} {output[19].ToString("0.000")}");
}
Console.ReadKey();
}
public NetworkState TrainNetwork()
{
int epoch = 0;
_trainingSet = new BasicNeuralDataSet(_annInputs, _annOutputs);
_basicNetwork = new BasicNetwork();
_basicNetwork.AddLayer(new BasicLayer(null, true, _nInputNeurons));
_basicNetwork.AddLayer(new BasicLayer(new ActivationSigmoid(), true, _nHiddenNeurons));
_basicNetwork.AddLayer(new BasicLayer(new ActivationSigmoid(), false, _nOutputNeurons));
_basicNetwork.Structure.FinalizeStructure();
_basicNetwork.Reset();
//Distribuisco numeri casuali[-1,1] lavorando in ambiente stocastico(non deterministico).
//In questo modo il training avviene in maniera casuale, partendo sempre dallo stesso stato.
new ConsistentRandomizer(-1, 1, 100).Randomize(_basicNetwork);
Backpropagation train = new Backpropagation(_basicNetwork, _trainingSet, LearnRate, Momentum);
train.FixFlatSpot = false;
do
{
train.Iteration();
epoch++;
_trainError = train.Error;
BufferTrainError.Add(_trainError);
} while (train.Error > _error);
train.FinishTraining();
_neuronsWeight = _basicNetwork.Structure.Network.Flat.Weights.Select(x => System.Convert.ToDouble(x)).ToList();
Make2DNeuronsWeightsMap();
foreach (IMLDataPair pair in _trainingSet)
{
IMLData output = _basicNetwork.Compute(pair.Input);
Console.WriteLine("Input: " + pair.Input[0] + @" - " + pair.Input[1] + @" - " + pair.Input[2]);
Console.WriteLine("Output 0: - actual=" + output[0] + @"-ideal=" + pair.Ideal[0]);
Console.WriteLine("Output 1: - actual=" + output[1] + @"-ideal=" + pair.Ideal[1]);
Console.WriteLine("Output 2: - actual=" + output[2] + @"-ideal=" + pair.Ideal[2]);
}
return(NetworkState.TRAINED);
}
/// <summary>
/// Measure the performance of the network
/// </summary>
/// <param name = "network">Network to analyze</param>
/// <param name = "dataset">Dataset with input and ideal data</param>
/// <returns>Error % of correct bits, returned by the network.</returns>
public static double MeasurePerformance(BasicNetwork network, BasicNeuralDataSet dataset)
{
int correctBits = 0;
float threshold;
IActivationFunction activationFunction = network.GetActivation(network.LayerCount - 1); // get the activation function of the output layer
if (activationFunction is ActivationSigmoid)
{
threshold = 0.5f; /* > 0.5, range of sigmoid [0..1]*/
}
else if (activationFunction is ActivationTANH)
{
threshold = 0.0f; /*> 0, range of bipolar sigmoid is [-1..1]*/
}
else
{
throw new ArgumentException("Bad activation function");
}
int n = (int)dataset.Count;
Parallel.For(
0,
n,
i =>
{
IMLData actualOutputs = network.Compute(dataset.Data[i].Input);
lock (LockObject)
{
for (int j = 0, k = actualOutputs.Count; j < k; j++)
{
if ((actualOutputs[j] > threshold && dataset.Data[i].Ideal[j] > threshold) ||
(actualOutputs[j] < threshold && dataset.Data[i].Ideal[j] < threshold))
{
correctBits++;
}
}
}
});
long totalOutputBitsCount = dataset.Count * dataset.Data[0].Ideal.Count;
return((double)correctBits / totalOutputBitsCount);
}
public static BasicNeuralDataSet LoadMonksData(string datasetLocation)
{
string[] lines = File.ReadAllLines(datasetLocation);
double[][] input = new double[lines.Length][];
double[][] output = new double[lines.Length][];
for (int i = 0; i < lines.Length; i++)
{
var line = lines[i].Trim().Split(' ');
input[i] = new double[17];
input[i][0] = line[1] == "1" ? 1 : 0;
input[i][1] = line[1] == "2" ? 1 : 0;
input[i][2] = line[1] == "3" ? 1 : 0;
input[i][3] = line[2] == "1" ? 1 : 0;
input[i][4] = line[2] == "2" ? 1 : 0;
input[i][5] = line[2] == "3" ? 1 : 0;
input[i][6] = line[3] == "1" ? 1 : 0;
input[i][7] = line[3] == "2" ? 1 : 0;
input[i][8] = line[4] == "1" ? 1 : 0;
input[i][9] = line[4] == "2" ? 1 : 0;
input[i][10] = line[4] == "3" ? 1 : 0;
input[i][11] = line[5] == "1" ? 1 : 0;
input[i][12] = line[5] == "2" ? 1 : 0;
input[i][13] = line[5] == "3" ? 1 : 0;
input[i][14] = line[5] == "4" ? 1 : 0;
input[i][15] = line[6] == "1" ? 1 : 0;
input[i][16] = line[6] == "2" ? 1 : 0;
output[i] = new double[] { double.Parse((line[0])) };
}
BasicNeuralDataSet trainingSet = new BasicNeuralDataSet(input, output);
return(trainingSet);
}
public void LoadCSVFileTrainingRegg()
{
var path = parameters.TrainFile;
var csvRead = new ReadCSV(new FileStream(path, FileMode.Open), true, CSVFormat.DecimalPoint);
var valuesX = new List <double[]>();
var valuesY = new List <double[]>();
while (csvRead.Next())
{
valuesX.Add(new double[1] {
csvRead.GetDouble(0)
});
valuesY.Add(new double[1] {
csvRead.GetDouble(1)
});
}
csvRead.Close();
var min = parameters.FunctionType == FunctionTypeEnum.Unipolar ? 0d : -1d;
ArgNormalize = new NormalizedField[] { new NormalizedField(NormalizationAction.Normalize, "X", valuesX.Max(v => v[0]), valuesX.Min(v => v[0]), 1.0, min) };
YNormalize = new NormalizedField(NormalizationAction.Normalize, "Y", valuesY.Max(v => v[0]), valuesY.Min(v => v[0]), 1.0, min);
for (int i = 0; i < valuesX.Count; i++)
{
valuesX[i][0] = ArgNormalize[0].Normalize(valuesX[i][0]);
valuesY[i][0] = YNormalize.Normalize(valuesY[i][0]);
}
//if(parameters.FunctionType==FunctionTypeEnum.Bipolar)
// ansMeans = means[0];
//ansfactor = factor[0];
//ansMIn = Min[0];
var trainSetCount = (int)((double)valuesX.Count * ((100.0 - 15) / 100));
valuesX.Shuffle();
valuesY.Shuffle();
MyExtensions.ResetStableShuffle();
TrainSet = new BasicNeuralDataSet(valuesX.Take(trainSetCount).ToArray(),
valuesY.Take(trainSetCount).ToArray());
ValidationSet = new BasicNeuralDataSet(valuesX.Skip(trainSetCount).ToArray(),
valuesY.Skip(trainSetCount).ToArray());
}
private BasicNeuralDataSet PrepareClassificationSet(double[][] data)
{
var input = data.Select(row => row.Take(_inputLayerSize).ToArray()).ToArray();
var idealTmp = data.Select(row => row.Skip(_inputLayerSize).ToArray()).ToArray();
var normalizedInput = _normalizer.Normalize(input);
var ideal = new double[idealTmp.Length][];
// Dzielimy wynik na klasy, "uaktywniając" odpowiadającą klasie kolumnę
for (int i = 0; i < idealTmp.Length; ++i)
{
ideal[i] = Enumerable.Repeat(GetNormalizationLowValue(), _outputLayerSize).ToArray();
ideal[i][(int)idealTmp[i][0] - 1] = 1;
}
var trainingSet = new BasicNeuralDataSet(normalizedInput, ideal);
return(trainingSet);
}
