private protected CrossValidationResult[] CrossValidateTrain(IDataView data, IEstimator <ITransformer> estimator,
int numFolds, string samplingKeyColumn, int?seed = null)
{
Environment.CheckValue(data, nameof(data));
Environment.CheckValue(estimator, nameof(estimator));
Environment.CheckParam(numFolds > 1, nameof(numFolds), "Must be more than 1");
Environment.CheckValueOrNull(samplingKeyColumn);
DataOperationsCatalog.EnsureGroupPreservationColumn(Environment, ref data, ref samplingKeyColumn, seed);
var result = new CrossValidationResult[numFolds];
int fold = 0;
// Sequential per-fold training.
// REVIEW: we could have a parallel implementation here. We would need to
// spawn off a separate host per fold in that case.
foreach (var split in DataOperationsCatalog.CrossValidationSplit(Environment, data, numFolds, samplingKeyColumn))
{
var model = estimator.Fit(split.TrainSet);
var scoredTest = model.Transform(split.TestSet);
result[fold] = new CrossValidationResult(model, scoredTest, fold);
fold++;
}
return(result);
}
public CrossValidationResult evaluate()
{
CrossValidationResult ret = new CrossValidationResult(modshogunPINVOKE.CrossValidation_evaluate(swigCPtr), true);
if (modshogunPINVOKE.SWIGPendingException.Pending)
{
throw modshogunPINVOKE.SWIGPendingException.Retrieve();
}
return(ret);
}
private void ClassifyDataByNaiveBayes(int numOfFolds = 3, int minOccurences = 1)
{
CalcInputAndOutputVariables(minOccurences);
var cvNaiveBayesClassifier = CrossValidation.Create(
k: numOfFolds,
learner: p => new NaiveBayesLearning <BernoulliDistribution>(),
loss: (actual, expected, p) => new ZeroOneLoss(expected).Loss(actual),
fit: (teacher, x, y, w) => teacher.Learn(x, y, w),
x: InputVariables,
y: OutputVariables
);
// Run Cross-Validation
Result = cvNaiveBayesClassifier.Learn(InputVariables, OutputVariables) as CrossValidationResult <TModel, double[], int>;
}
public static void saveResultsToFile(string file, CrossValidationResult[] results)
{
var stream = File.Open(file, FileMode.Create, FileAccess.Write);
var sw = new StreamWriter(stream);
sw.WriteLine(CrossValidationResult.getCSVHead());
foreach (var result in results)
{
foreach (var row in result.getCSVData())
{
sw.WriteLine(row);
}
}
sw.Close();
}
public static void Main(string[] args)
{
// create data set from csv file
MultiLayerPerceptron neuralNet = (MultiLayerPerceptron)NeuralNetwork.createFromFile("irisNet.nnet");
DataSet dataSet = DataSet.createFromFile("data_sets/iris_data_normalised.txt", 4, 3, ",");
string[] classNames = new string[] { "Virginica", "Setosa", "Versicolor" };
CrossValidation crossval = new CrossValidation(neuralNet, dataSet, 5);
crossval.addEvaluator(new ClassifierEvaluator.MultiClass(classNames));
crossval.run();
CrossValidationResult results = crossval.Result;
Console.WriteLine(results);
}
/// <summary>
/// Train the <paramref name="estimator"/> on <paramref name="numFolds"/> folds of the data sequentially.
/// Return each model and each scored test dataset.
/// </summary>
protected internal CrossValidationResult[] CrossValidateTrain(IDataView data, IEstimator <ITransformer> estimator,
int numFolds, string samplingKeyColumn, uint?seed = null)
{
Environment.CheckValue(data, nameof(data));
Environment.CheckValue(estimator, nameof(estimator));
Environment.CheckParam(numFolds > 1, nameof(numFolds), "Must be more than 1");
Environment.CheckValueOrNull(samplingKeyColumn);
EnsureGroupPreservationColumn(ref data, ref samplingKeyColumn, seed);
Func <int, CrossValidationResult> foldFunction =
fold =>
{
var trainFilter = new RangeFilter(Environment, new RangeFilter.Options
{
Column = samplingKeyColumn,
Min = (double)fold / numFolds,
Max = (double)(fold + 1) / numFolds,
Complement = true
}, data);
var testFilter = new RangeFilter(Environment, new RangeFilter.Options
{
Column = samplingKeyColumn,
Min = (double)fold / numFolds,
Max = (double)(fold + 1) / numFolds,
Complement = false
}, data);
var model = estimator.Fit(trainFilter);
var scoredTest = model.Transform(testFilter);
return(new CrossValidationResult(model, scoredTest, fold));
};
// Sequential per-fold training.
// REVIEW: we could have a parallel implementation here. We would need to
// spawn off a separate host per fold in that case.
var result = new CrossValidationResult[numFolds];
for (int fold = 0; fold < numFolds; fold++)
{
result[fold] = foldFunction(fold);
}
return(result);
}
private void ClassifyDataByLogisticRegression(int numOfFolds = 3, int minOccurences = 1, int maxIterations = 100)
{
CalcInputAndOutputVariables(minOccurences);
var cvLogisticRegressionClassifier = CrossValidation.Create(
k: numOfFolds,
learner: (p) => new IterativeReweightedLeastSquares <LogisticRegression>()
{
MaxIterations = 100,
Regularization = 1e-6
},
loss: (actual, expected, p) => new ZeroOneLoss(expected).Loss(actual),
fit: (teacher, x, y, w) => teacher.Learn(x, y, w),
x: InputVariables,
y: OutputVariables
);
// Run Cross-Validation
Result = cvLogisticRegressionClassifier.Learn(InputVariables, OutputVariables) as CrossValidationResult <TModel, double[], int>;
}
internal static HandleRef getCPtr(CrossValidationResult obj) {
return (obj == null) ? new HandleRef(null, IntPtr.Zero) : obj.swigCPtr;
}
internal static HandleRef getCPtr(CrossValidationResult obj)
{
return((obj == null) ? new HandleRef(null, IntPtr.Zero) : obj.swigCPtr);
}
public CrossValidationResult evaluate() {
CrossValidationResult ret = new CrossValidationResult(modshogunPINVOKE.CrossValidation_evaluate(swigCPtr), true);
if (modshogunPINVOKE.SWIGPendingException.Pending) throw modshogunPINVOKE.SWIGPendingException.Retrieve();
return ret;
}
