/// <inheritdoc />
public String SuggestModelArchitecture(VersatileMLDataSet dataset)
{
int inputColumns = dataset.NormHelper.InputColumns.Count;
int outputColumns = dataset.NormHelper.OutputColumns.Count;
var hiddenCount = (int) ((inputColumns + outputColumns)*1.5);
var result = new StringBuilder();
result.Append("?->gaussian(c=");
result.Append(hiddenCount);
result.Append(")->?");
return result.ToString();
}
/// <inheritdoc />
public INormalizationStrategy SuggestNormalizationStrategy(VersatileMLDataSet dataset, String architecture)
{
var result = new BasicNormalizationStrategy();
result.AssignInputNormalizer(ColumnType.Continuous, new RangeNormalizer(0, 1));
result.AssignInputNormalizer(ColumnType.Nominal, new OneOfNNormalizer(0, 1));
result.AssignInputNormalizer(ColumnType.Ordinal, new OneOfNNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Continuous, new RangeNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Nominal, new OneOfNNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Ordinal, new OneOfNNormalizer(0, 1));
return result;
}
public void TestNOrmalizationHelperSerialization()
{
try
{
var trainingData = new VersatileMLDataSet();
var nh = trainingData.NormHelper;
SerializeRoundTrip.RoundTrip(nh);
}
catch (Exception ex)
{
Assert.Fail("Error in TestNomalizationHelperSerialization: {0}", ex);
}
}
/// <inheritdoc />
public String SuggestModelArchitecture(VersatileMLDataSet dataset)
{
int outputColumns = dataset.NormHelper.OutputColumns.Count;
if (outputColumns > 1)
{
throw new EncogError("SVM does not support multiple output columns.");
}
ColumnType ct = dataset.NormHelper.OutputColumns[0].DataType;
var result = new StringBuilder();
result.Append("?->");
result.Append(ct == ColumnType.Nominal ? "C" : "R");
result.Append("->?");
return result.ToString();
}
/// <inheritdoc />
public INormalizationStrategy SuggestNormalizationStrategy(VersatileMLDataSet dataset, string architecture)
{
int outputColumns = dataset.NormHelper.OutputColumns.Count;
ColumnType ct = dataset.NormHelper.OutputColumns[0].DataType;
var result = new BasicNormalizationStrategy();
result.AssignInputNormalizer(ColumnType.Continuous, new RangeNormalizer(0, 1));
result.AssignInputNormalizer(ColumnType.Nominal, new OneOfNNormalizer(0, 1));
result.AssignInputNormalizer(ColumnType.Ordinal, new OneOfNNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Continuous, new RangeNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Nominal, new OneOfNNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Ordinal, new OneOfNNormalizer(0, 1));
return result;
}
static void Main(string[] args)
{
var format = new CSVFormat('.', ',');
IVersatileDataSource source = new CSVDataSource("input.csv", true, format);
var data = new VersatileMLDataSet(source);
data.NormHelper.Format = format;
for (int i = 0; i < 228; i++)
{
data.DefineSourceColumn("a" + (i + 1), i, ColumnType.Continuous);
}
ColumnDefinition columnOutput = data.DefineSourceColumn("close", 228, ColumnType.Continuous);
data.Analyze();
data.DefineSingleOutputOthersInput(columnOutput);
var model = new EncogModel(data);
model.SelectMethod(data, MLMethodFactory.TypeFeedforward);
model.Report = new ConsoleStatusReportable();
data.LeadWindowSize = 1;
data.LagWindowSize = 2;
data.Normalize();
model.HoldBackValidation(0.3, true, 1001);
model.SelectTrainingType(data);
// Use a 5-fold cross-validated train. Return the best method found.
var bestMethod = (IMLRegression)model.Crossvalidate(5, true);
// Display the training and validation errors.
Console.WriteLine(@"Training error: " + model.CalculateError(bestMethod, model.TrainingDataset));
Console.WriteLine(@"Validation error: " + model.CalculateError(bestMethod, model.ValidationDataset));
// Display our normalization parameters.
NormalizationHelper helper = data.NormHelper;
Console.WriteLine(helper.ToString());
// Display the final model.
Console.WriteLine(@"Final model: " + bestMethod);
}
/// <inheritdoc />
public INormalizationStrategy SuggestNormalizationStrategy(VersatileMLDataSet dataset, String architecture)
{
int outputColumns = dataset.NormHelper.OutputColumns.Count;
if (outputColumns > 1)
{
throw new EncogError("PNN does not support multiple output columns.");
}
ColumnType ct = dataset.NormHelper.OutputColumns[0].DataType;
var result = new BasicNormalizationStrategy();
result.AssignInputNormalizer(ColumnType.Continuous, new RangeNormalizer(0, 1));
result.AssignInputNormalizer(ColumnType.Nominal, new OneOfNNormalizer(0, 1));
result.AssignInputNormalizer(ColumnType.Ordinal, new OneOfNNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Continuous, new RangeNormalizer(0, 1));
result.AssignOutputNormalizer(ColumnType.Nominal, new IndexedNormalizer());
result.AssignOutputNormalizer(ColumnType.Ordinal, new OneOfNNormalizer(0, 1));
return result;
}
/// <inheritdoc />
public String SuggestModelArchitecture(VersatileMLDataSet dataset)
{
return ("cycles=4");
}
/// <summary>
/// Program entry point.
/// </summary>
/// <param name="app">Holds arguments and other info.</param>
public void Execute(IExampleInterface app)
{
// Download the data that we will attempt to model.
string filename = DownloadData(app.Args);
// Define the format of the data file.
// This area will change, depending on the columns and
// format of the file that you are trying to model.
var format = new CSVFormat('.', ' '); // decimal point and space separated
IVersatileDataSource source = new CSVDataSource(filename, false, format);
var data = new VersatileMLDataSet(source);
data.NormHelper.Format = format;
ColumnDefinition columnMPG = data.DefineSourceColumn("mpg", 0, ColumnType.Continuous);
ColumnDefinition columnCylinders = data.DefineSourceColumn("cylinders", 1, ColumnType.Ordinal);
// It is very important to predefine ordinals, so that the order is known.
columnCylinders.DefineClass(new[] {"3", "4", "5", "6", "8"});
data.DefineSourceColumn("displacement", 2, ColumnType.Continuous);
ColumnDefinition columnHorsePower = data.DefineSourceColumn("horsepower", 3, ColumnType.Continuous);
data.DefineSourceColumn("weight", 4, ColumnType.Continuous);
data.DefineSourceColumn("acceleration", 5, ColumnType.Continuous);
ColumnDefinition columnModelYear = data.DefineSourceColumn("model_year", 6, ColumnType.Ordinal);
columnModelYear.DefineClass(new[]
{"70", "71", "72", "73", "74", "75", "76", "77", "78", "79", "80", "81", "82"});
data.DefineSourceColumn("origin", 7, ColumnType.Nominal);
// Define how missing values are represented.
data.NormHelper.DefineUnknownValue("?");
data.NormHelper.DefineMissingHandler(columnHorsePower, new MeanMissingHandler());
// Analyze the data, determine the min/max/mean/sd of every column.
data.Analyze();
// Map the prediction column to the output of the model, and all
// other columns to the input.
data.DefineSingleOutputOthersInput(columnMPG);
// Create feedforward neural network as the model type. MLMethodFactory.TYPE_FEEDFORWARD.
// You could also other model types, such as:
// MLMethodFactory.SVM: Support Vector Machine (SVM)
// MLMethodFactory.TYPE_RBFNETWORK: RBF Neural Network
// MLMethodFactor.TYPE_NEAT: NEAT Neural Network
// MLMethodFactor.TYPE_PNN: Probabilistic Neural Network
var model = new EncogModel(data);
model.SelectMethod(data, MLMethodFactory.TypeFeedforward);
// Send any output to the console.
model.Report = new ConsoleStatusReportable();
// Now normalize the data. Encog will automatically determine the correct normalization
// type based on the model you chose in the last step.
data.Normalize();
// Hold back some data for a final validation.
// Shuffle the data into a random ordering.
// Use a seed of 1001 so that we always use the same holdback and will get more consistent results.
model.HoldBackValidation(0.3, true, 1001);
// Choose whatever is the default training type for this model.
model.SelectTrainingType(data);
// Use a 5-fold cross-validated train. Return the best method found.
var bestMethod = (IMLRegression) model.Crossvalidate(5, true);
// Display the training and validation errors.
Console.WriteLine(@"Training error: " + model.CalculateError(bestMethod, model.TrainingDataset));
Console.WriteLine(@"Validation error: " + model.CalculateError(bestMethod, model.ValidationDataset));
// Display our normalization parameters.
NormalizationHelper helper = data.NormHelper;
Console.WriteLine(helper.ToString());
// Display the final model.
Console.WriteLine("Final model: " + bestMethod);
// Loop over the entire, original, dataset and feed it through the model.
// This also shows how you would process new data, that was not part of your
// training set. You do not need to retrain, simply use the NormalizationHelper
// class. After you train, you can save the NormalizationHelper to later
// normalize and denormalize your data.
source.Close();
var csv = new ReadCSV(filename, false, format);
var line = new String[7];
IMLData input = helper.AllocateInputVector();
while (csv.Next())
{
var result = new StringBuilder();
line[0] = csv.Get(1);
line[1] = csv.Get(2);
line[2] = csv.Get(3);
line[3] = csv.Get(4);
line[4] = csv.Get(5);
line[5] = csv.Get(6);
line[6] = csv.Get(7);
String correct = csv.Get(0);
helper.NormalizeInputVector(line, ((BasicMLData) input).Data, false);
IMLData output = bestMethod.Compute(input);
String irisChosen = helper.DenormalizeOutputVectorToString(output)[0];
result.Append(line);
result.Append(" -> predicted: ");
result.Append(irisChosen);
result.Append("(correct: ");
result.Append(correct);
result.Append(")");
Console.WriteLine(result.ToString());
}
csv.Close();
// Delete data file and shut down.
File.Delete(filename);
EncogFramework.Instance.Shutdown();
}
/// <summary>
/// Program entry point.
/// </summary>
/// <param name="app">Holds arguments and other info.</param>
public void Execute(IExampleInterface app)
{
ErrorCalculation.Mode = ErrorCalculationMode.RMS;
// Download the data that we will attempt to model.
string filename = DownloadData(app.Args);
// Define the format of the data file.
// This area will change, depending on the columns and
// format of the file that you are trying to model.
var format = new CSVFormat('.', ' '); // decimal point and
// space separated
IVersatileDataSource source = new CSVDataSource(filename, true,
format);
var data = new VersatileMLDataSet(source);
data.NormHelper.Format = format;
ColumnDefinition columnSSN = data.DefineSourceColumn("SSN",
ColumnType.Continuous);
ColumnDefinition columnDEV = data.DefineSourceColumn("DEV",
ColumnType.Continuous);
// Analyze the data, determine the min/max/mean/sd of every column.
data.Analyze();
// Use SSN & DEV to predict SSN. For time-series it is okay to have
// SSN both as
// an input and an output.
data.DefineInput(columnSSN);
data.DefineInput(columnDEV);
data.DefineOutput(columnSSN);
// Create feedforward neural network as the model type.
// MLMethodFactory.TYPE_FEEDFORWARD.
// You could also other model types, such as:
// MLMethodFactory.SVM: Support Vector Machine (SVM)
// MLMethodFactory.TYPE_RBFNETWORK: RBF Neural Network
// MLMethodFactor.TYPE_NEAT: NEAT Neural Network
// MLMethodFactor.TYPE_PNN: Probabilistic Neural Network
var model = new EncogModel(data);
model.SelectMethod(data, MLMethodFactory.TypeFeedforward);
// Send any output to the console.
model.Report = new ConsoleStatusReportable();
// Now normalize the data. Encog will automatically determine the
// correct normalization
// type based on the model you chose in the last step.
data.Normalize();
// Set time series.
data.LeadWindowSize = 1;
data.LagWindowSize = WindowSize;
// Hold back some data for a final validation.
// Do not shuffle the data into a random ordering. (never shuffle
// time series)
// Use a seed of 1001 so that we always use the same holdback and
// will get more consistent results.
model.HoldBackValidation(0.3, false, 1001);
// Choose whatever is the default training type for this model.
model.SelectTrainingType(data);
// Use a 5-fold cross-validated train. Return the best method found.
// (never shuffle time series)
var bestMethod = (IMLRegression) model.Crossvalidate(5,
false);
// Display the training and validation errors.
Console.WriteLine(@"Training error: "
+ model.CalculateError(bestMethod,
model.TrainingDataset));
Console.WriteLine(@"Validation error: "
+ model.CalculateError(bestMethod,
model.ValidationDataset));
// Display our normalization parameters.
NormalizationHelper helper = data.NormHelper;
Console.WriteLine(helper.ToString());
// Display the final model.
Console.WriteLine(@"Final model: " + bestMethod);
// Loop over the entire, original, dataset and feed it through the
// model. This also shows how you would process new data, that was
// not part of your training set. You do not need to retrain, simply
// use the NormalizationHelper class. After you train, you can save
// the NormalizationHelper to later normalize and denormalize your
// data.
source.Close();
var csv = new ReadCSV(filename, true, format);
var line = new String[2];
// Create a vector to hold each time-slice, as we build them.
// These will be grouped together into windows.
var slice = new double[2];
var window = new VectorWindow(WindowSize + 1);
IMLData input = helper.AllocateInputVector(WindowSize + 1);
// Only display the first 100
int stopAfter = 100;
while (csv.Next() && stopAfter > 0)
{
var result = new StringBuilder();
line[0] = csv.Get(2); // ssn
line[1] = csv.Get(3); // dev
helper.NormalizeInputVector(line, slice, false);
// enough data to build a full window?
if (window.IsReady())
{
window.CopyWindow(((BasicMLData) input).Data, 0);
String correct = csv.Get(2); // trying to predict SSN.
IMLData output = bestMethod.Compute(input);
String predicted = helper
.DenormalizeOutputVectorToString(output)[0];
result.Append(line);
result.Append(" -> predicted: ");
result.Append(predicted);
result.Append("(correct: ");
result.Append(correct);
result.Append(")");
Console.WriteLine(result.ToString());
}
// Add the normalized slice to the window. We do this just after
// the after checking to see if the window is ready so that the
// window is always one behind the current row. This is because
// we are trying to predict next row.
window.Add(slice);
stopAfter--;
}
csv.Close();
// Delete data file and shut down.
File.Delete(filename);
EncogFramework.Instance.Shutdown();
}
/// <inheritdoc />
public int DetermineOutputCount(VersatileMLDataSet dataset)
{
return dataset.NormHelper.OutputColumns[0].Classes.Count;
}
/// <inheritdoc />
public String SuggestModelArchitecture(VersatileMLDataSet dataset)
{
return ("?->C(kernel=gaussian)->?");
}
/// <summary>
/// Select the training type.
/// </summary>
/// <param name="dataset">The dataset.</param>
public void SelectTrainingType(VersatileMLDataSet dataset)
{
if (_methodType == null)
{
throw new EncogError(
"Please select your training method, before your training type.");
}
IMethodConfig config = _methodConfigurations[_methodType];
SelectTraining(dataset, config.SuggestTrainingType(),
config.SuggestTrainingArgs(_trainingType));
}
/// <summary>
/// Select the method to create.
/// </summary>
/// <param name="dataset">The dataset.</param>
/// <param name="methodType">The method type.</param>
public void SelectMethod(VersatileMLDataSet dataset, String methodType)
{
if (!_methodConfigurations.ContainsKey(methodType))
{
throw new EncogError("Don't know how to autoconfig method: "
+ methodType);
}
_config = _methodConfigurations[methodType];
_methodType = methodType;
_methodArgs = _config.SuggestModelArchitecture(dataset);
dataset.NormHelper.NormStrategy =
_config.SuggestNormalizationStrategy(dataset, _methodArgs);
}
/// <summary>
/// Select the method to use.
/// </summary>
/// <param name="dataset">The dataset.</param>
/// <param name="methodType">The type of method.</param>
/// <param name="methodArgs">The method arguments.</param>
/// <param name="trainingType">The training type.</param>
/// <param name="trainingArgs">The training arguments.</param>
public void SelectMethod(VersatileMLDataSet dataset, String methodType,
String methodArgs, String trainingType, String trainingArgs)
{
if (!_methodConfigurations.ContainsKey(methodType))
{
throw new EncogError("Don't know how to autoconfig method: "
+ methodType);
}
_methodType = methodType;
_methodArgs = methodArgs;
dataset.NormHelper.NormStrategy =
_methodConfigurations[methodType]
.SuggestNormalizationStrategy(dataset, methodArgs);
}
/// <summary>
/// Construct a model for the specified dataset.
/// </summary>
/// <param name="theDataset">The dataset.</param>
public EncogModel(VersatileMLDataSet theDataset) : this()
{
_dataset = theDataset;
_methodConfigurations[MLMethodFactory.TypeFeedforward] = new FeedforwardConfig();
_methodConfigurations[MLMethodFactory.TypeSVM] = new SVMConfig();
_methodConfigurations[MLMethodFactory.TypeRbfnetwork] = new RBFNetworkConfig();
_methodConfigurations[MLMethodFactory.TypeNEAT] = new NEATConfig();
_methodConfigurations[MLMethodFactory.TypePNN] = new PNNConfig();
}
/// <inheritdoc />
public int DetermineOutputCount(VersatileMLDataSet dataset)
{
return dataset.NormHelper.CalculateNormalizedOutputCount();
}
/// <summary>
/// Select the training to use.
/// </summary>
/// <param name="dataset">The dataset.</param>
/// <param name="trainingType">The type of training.</param>
/// <param name="trainingArgs">The training arguments.</param>
public void SelectTraining(VersatileMLDataSet dataset, String trainingType,
String trainingArgs)
{
if (_methodType == null)
{
throw new EncogError(
"Please select your training method, before your training type.");
}
_trainingType = trainingType;
_trainingArgs = trainingArgs;
}