public void RegressionNeuralNetModel_Save()
{
var numberOfObservations = 500;
var numberOfFeatures = 5;
var random = new Random(32);
var observations = new F64Matrix(numberOfObservations, numberOfFeatures);
observations.Map(() => random.NextDouble());
var targets = Enumerable.Range(0, numberOfObservations)
.Select(i => (double)random.NextDouble()).ToArray();
var net = new NeuralNet();
net.Add(new InputLayer(numberOfFeatures));
net.Add(new DenseLayer(10));
net.Add(new SquaredErrorRegressionLayer());
var learner = new RegressionNeuralNetLearner(net, new AccuracyLoss());
var sut = learner.Learn(observations, targets);
// save model.
var writer = new StringWriter();
sut.Save(() => writer);
// load model and assert prediction results.
sut = RegressionNeuralNetModel.Load(() => new StringReader(writer.ToString()));
var predictions = sut.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.0871547675024143, actual, 0.0001);
}
public void RegressionNeuralNetLearner_Learn_Early_Stopping()
{
var numberOfObservations = 500;
var numberOfFeatures = 5;
var random = new Random(32);
F64Matrix observations;
double[] targets;
CreateData(numberOfObservations, numberOfFeatures, random, out observations, out targets);
F64Matrix validationObservations;
double[] validationTargets;
CreateData(numberOfObservations, numberOfFeatures, random, out validationObservations, out validationTargets);
var net = new NeuralNet();
net.Add(new InputLayer(numberOfFeatures));
net.Add(new DenseLayer(10));
net.Add(new SquaredErrorRegressionLayer());
var sut = new RegressionNeuralNetLearner(net, new SquareLoss(), 0.01, 150);
var model = sut.Learn(observations, targets,
validationObservations, validationTargets);
var validationPredictions = model.Predict(validationObservations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(validationTargets, validationPredictions);
Assert.AreEqual(0.093500629562319859, actual, 0.0001);
}
public void SharpLearning_Example()
{
// Load data
var(observations, targets) = DataSetUtilities.LoadWinequalityWhite();
// transform data for neural net
var transform = new MinMaxTransformer(0.0, 1.0);
transform.Transform(observations, observations);
var featureCount = observations.ColumnCount;
// define the neural net.
var net = new NeuralNet();
net.Add(new InputLayer(inputUnits: featureCount));
net.Add(new DenseLayer(32, Activation.Relu));
net.Add(new DenseLayer(32, Activation.Relu));
net.Add(new SquaredErrorRegressionLayer());
// using only 10 iteration to make the example run faster.
// using square error as error metric. This is only used for reporting progress.
var learner = new RegressionNeuralNetLearner(net, iterations: 10, loss: new SquareLoss(),
optimizerMethod: OptimizerMethod.Adam);
var cv = new RandomCrossValidation <double>(10, seed: 232);
var predictions = cv.CrossValidate(learner, observations, targets);
Trace.WriteLine(FormatErrorString(targets, predictions));
}
public void RegressionNeuralNetLearner_Learn()
{
var numberOfObservations = 500;
var numberOfFeatures = 5;
var random = new Random(32);
F64Matrix observations;
double[] targets;
CreateData(numberOfObservations, numberOfFeatures, random, out observations, out targets);
var net = new NeuralNet();
net.Add(new InputLayer(numberOfFeatures));
net.Add(new DenseLayer(10));
net.Add(new SquaredErrorRegressionLayer());
var sut = new RegressionNeuralNetLearner(net, new SquareLoss());
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.0871547675024143, actual, 0.0001);
}
// Start is called before the first frame update
void Start()
{
net = new NeuralNet();
net.Add(new InputLayer(2));
net.Add(new DenseLayer(8));
net.Add(new DenseLayer(8));
net.Add(new SquaredErrorRegressionLayer());
learner = new RegressionNeuralNetLearner(net, new SquareLoss(), iterations: 1, batchSize: 1);
}
public void RegressionNeuralNetLearner_Constructor_Throw_On_Wrong_OutputLayerType()
{
var net = new NeuralNet();
net.Add(new InputLayer(10));
net.Add(new DenseLayer(10));
net.Add(new SvmLayer(10));
var sut = new RegressionNeuralNetLearner(net, new AccuracyLoss());
}
private ILearner <double> GetNeuralnet(int numberOfFeatures)
{
var net = new NeuralNet();
net.Add(new InputLayer(inputUnits: numberOfFeatures));
net.Add(new DenseLayer(numberOfFeatures, Activation.Relu));
net.Add(new SquaredErrorRegressionLayer());
var learner = new RegressionNeuralNetLearner(net, learningRate: 0.001, iterations: 2000, loss: new SquareLoss(), batchSize: 180, optimizerMethod: OptimizerMethod.Adam);
return(learner);
}
public void FeatureNormalization_Normalize()
{
// Use StreamReader(filepath) when running from filesystem
var parser = new CsvParser(() => new StringReader(Resources.winequality_white));
var targetName = "quality";
// read feature matrix (all columns different from the targetName)
var observations = parser.EnumerateRows(c => c != targetName)
.ToF64Matrix();
// create minmax normalizer (normalizes each feature from 0.0 to 1.0)
var minMaxTransformer = new MinMaxTransformer(0.0, 1.0);
// transforms features using the feature normalization transform
minMaxTransformer.Transform(observations, observations);
// read targets
var targets = parser.EnumerateRows(targetName)
.ToF64Vector();
// Create neural net.
var net = new NeuralNet();
net.Add(new InputLayer(observations.ColumnCount));
net.Add(new SquaredErrorRegressionLayer());
// Create regression learner.
var learner = new RegressionNeuralNetLearner(net, new SquareLoss());
// learns a neural net regression model.
var model = learner.Learn(observations, targets);
// serializer for saving the MinMaxTransformer
var serializer = new GenericXmlDataContractSerializer();
// Serialize transform for use with the model.
// Replace this with StreamWriter for use with file system.
var data = new StringBuilder();
var writer = new StringWriter(data);
serializer.Serialize(minMaxTransformer, () => writer);
// Deserialize transform for use with the model.
// Replace this with StreamReader for use with file system.
var reader = new StringReader(data.ToString());
var deserializedMinMaxTransform = serializer.Deserialize <MinMaxTransformer>(() => reader);
// Normalize observation and predict using the model.
var normalizedObservation = deserializedMinMaxTransform.Transform(observations.Row(0));
var prediction = model.Predict(normalizedObservation);
Trace.WriteLine($"Prediction: {prediction}");
}
public void Regression_Standard_Neural_Net_FeatureTransform_Normalization()
{
#region Read Data
// Use StreamReader(filepath) when running from filesystem
var parser = new CsvParser(() => new StringReader(Resources.winequality_white));
var targetName = "quality";
// read feature matrix
var observations = parser.EnumerateRows(c => c != targetName)
.ToF64Matrix();
// read classification targets
var targets = parser.EnumerateRows(targetName)
.ToF64Vector();
#endregion
// transform pixel values to be between 0 and 1
// and shift each feature to have a mean value of zero.
var minMaxTransformer = new MinMaxTransformer(0.0, 1.0);
var meanZeroTransformer = new MeanZeroFeatureTransformer();
minMaxTransformer.Transform(observations, observations);
meanZeroTransformer.Transform(observations, observations);
var numberOfFeatures = observations.ColumnCount;
// define the neural net.
var net = new NeuralNet();
net.Add(new InputLayer(inputUnits: numberOfFeatures));
net.Add(new DropoutLayer(0.2));
net.Add(new DenseLayer(800, Activation.Relu));
net.Add(new DropoutLayer(0.5));
net.Add(new DenseLayer(800, Activation.Relu));
net.Add(new DropoutLayer(0.5));
net.Add(new SquaredErrorRegressionLayer());
// using only 10 iteration to make the example run faster.
// using square error as error metric. This is only used for reporting progress.
var learner = new RegressionNeuralNetLearner(net, iterations: 10, loss: new SquareLoss());
var model = learner.Learn(observations, targets);
var metric = new MeanSquaredErrorRegressionMetric();
var predictions = model.Predict(observations);
Trace.WriteLine("Training Error: " + metric.Error(targets, predictions));
}
private ILearner <double> GetNeuralNet(int numberOfFeatures, int batchSize, int?iterations = null)
{
var net = new NeuralNet();
net.Add(new InputLayer(inputUnits: numberOfFeatures));
net.Add(new DenseLayer(numberOfFeatures, Activation.Relu)
{
BatchNormalization = true
});
net.Add(new SquaredErrorRegressionLayer());
var learner = new RegressionNeuralNetLearner(net, learningRate: 0.01, iterations: iterations ?? _iterations, loss: new SquareLoss(), batchSize: batchSize,
optimizerMethod: OptimizerMethod.RMSProp);
return(learner);
}
/// <summary>
/// Predição de Floresta Aleatória e Rede Neural
/// </summary>
public void RegressionLearner_Learn_And_Predict()
{
#region Treinamento da Floresta Aleatória
var parser = new CsvParser(() => new StringReader(treinamento));
var targetName = "T";
var observations = parser.EnumerateRows(c => c != targetName)
.ToF64Matrix();
var targets = parser.EnumerateRows(targetName)
.ToF64Vector();
UltimaObservacao = new double[] { observations[observations.RowCount - 1, 0], observations[observations.RowCount - 1, 2], observations[observations.RowCount - 1, 3], observations[observations.RowCount - 1, 4], observations[observations.RowCount - 1, 5], targets[targets.Count() - 1] };
var learner = new RegressionRandomForestLearner(trees: 500);
model = learner.Learn(observations, targets);
#endregion
#region Teste da Floresta Aleatória
parser = new CsvParser(() => new StringReader(teste));
var observationsTeste = parser.EnumerateRows(c => c != targetName)
.ToF64Matrix();
var targetsTeste = parser.EnumerateRows(targetName)
.ToF64Vector();
// predict the training and test set.
var trainPredictions = model.Predict(observations);
var testPredictions = model.Predict(observationsTeste);
// create the metric
var metric = new MeanSquaredErrorRegressionMetric();
// measure the error on training and test set.
trainError = metric.Error(targets, trainPredictions);
testError = metric.Error(targetsTeste, testPredictions);
#endregion
#region Treinamento da Rede Neural
var net = new NeuralNet();
net.Add(new InputLayer(6));
net.Add(new DropoutLayer(0.2));
net.Add(new DenseLayer(800, Activation.Relu));
net.Add(new DropoutLayer(0.5));
net.Add(new DenseLayer(800, Activation.Relu));
net.Add(new DropoutLayer(0.5));
net.Add(new SquaredErrorRegressionLayer());
var learnernet = new RegressionNeuralNetLearner(net, iterations: 500, loss: new SquareLoss());
modelnet = learnernet.Learn(observations, targets);
#endregion
#region Teste da Rede Neural
trainPredictions = modelnet.Predict(observations);
testPredictions = modelnet.Predict(observationsTeste);
trainErrorNet = metric.Error(targets, trainPredictions);
testErrorNet = metric.Error(targetsTeste, testPredictions);
#endregion
#region Treinamento Ada
var learnerada = new RegressionAdaBoostLearner(maximumTreeDepth: 35, iterations: 2000, learningRate: 0.1);
modelada = learnerada.Learn(observations, targets);
#endregion
#region Teste Ada
trainPredictions = modelada.Predict(observations);
testPredictions = modelada.Predict(observationsTeste);
trainErrorAda = metric.Error(targets, trainPredictions);
testErrorAda = metric.Error(targetsTeste, testPredictions);
string stargets = "";
string strainPredictions = "";
string stargetsTeste = "";
string stestPredictions = "";
foreach (var i in targets)
{
stargets += i + ";";
}
foreach (var i in trainPredictions)
{
strainPredictions += i + ";";
}
foreach (var i in targetsTeste)
{
stargetsTeste += i + ";";
}
foreach (var i in testPredictions)
{
stestPredictions += i + ";";
}
#endregion
}
static void Main(string[] args)
{
string[] races = { "Black", "Hispanic", "NativeAmerican", "White" };
foreach (string race in races)
{
var dataTable = Utilities.GetDataTable(
$@"SELECT [StateId]
,[Year]
,[Subsidy]
,[Married]
,[AverageAge]
,[Male]
,[{race}]
,[NonRelative]
,[AverageMonths]
FROM[AdoptionStatistics].[dbo].[ObservationsByStateYear]", connectionString);
// do one hot encoding before we split the table up into training and test
// otherwise we will have a problem if the StateId column has different number of distinct values,
// which it actually does for 2016: data for Puerto Rico 2016 is not available
dataTable = Utilities.OneHotEncoder(dataTable, new string[] { "StateId" });
// get training data, all years except most recent, 2016
var trainingRows = dataTable.Select("[Year] <> 2016");
// get most recent year, 2016, for evaluating model
var testRows = dataTable.Select("[Year] = 2016");
(var observationsTraining, var targetsTraining) = Utilities.ObservationsFromDataRows(trainingRows, "AverageMonths");
(var observationsTest, var targetsTest) = Utilities.ObservationsFromDataRows(testRows, "AverageMonths");
// create neural net
var adoptStatistics = new NeuralNet();
adoptStatistics.Add(new InputLayer(inputUnits: dataTable.Columns.Count - 1));
adoptStatistics.Add(new DenseLayer(1200, Activation.Relu));
adoptStatistics.Add(new DenseLayer(1200, Activation.Relu));
adoptStatistics.Add(new DenseLayer(1200, Activation.Relu));
adoptStatistics.Add(new DenseLayer(1200, Activation.Relu));
adoptStatistics.Add(new SquaredErrorRegressionLayer());
// create learner, set iterations
var learner = new RegressionNeuralNetLearner(adoptStatistics, iterations: 1000, loss: new SquareLoss());
var model = learner.Learn(observationsTraining, targetsTraining);
DirectoryInfo directory = new DirectoryInfo(outputDirectory);
if (!directory.Exists)
{
directory.Create();
}
string modelName = $"adoptModel_{race}_{DateTime.Now:yy.MM.dd.HH.mm.ss}";
model.Save(() => new StreamWriter(Path.Combine(directory.FullName, $"{modelName}.xml")));
var metric = new MeanSquaredErrorRegressionMetric();
var predictionsTraining = model.Predict(observationsTraining);
var errorTraining = metric.Error(targetsTraining, predictionsTraining);
Console.WriteLine($"Training Error for {modelName}: {errorTraining}");
var predictionsTest = model.Predict(observationsTest);
var errorTest = metric.Error(targetsTest, predictionsTest);
Console.WriteLine($"Evaluation Error for {modelName}: {errorTest}");
}
}
