public RunCollectionChartAggregationView() {
InitializeComponent();
runMapping = new Dictionary<IRun, IEnumerable<DataRow>>();
combinedDataTable = new DataTable("Combined DataTable", "A data table containing data rows from multiple runs.");
viewHost.Content = combinedDataTable;
suppressUpdates = false;
}
public override IOperation Apply() {
var results = ResultCollectionParameter.ActualValue;
if (!results.ContainsKey(RegressionSolutionQualitiesResultName)) {
var newDataTable = new DataTable(RegressionSolutionQualitiesResultName);
results.Add(new Result(RegressionSolutionQualitiesResultName, "Chart displaying the training and test qualities of the regression solutions.", newDataTable));
}
var dataTable = (DataTable)results[RegressionSolutionQualitiesResultName].Value;
foreach (var result in results.Where(r => r.Value is IRegressionSolution)) {
var solution = (IRegressionSolution)result.Value;
var trainingR2 = result.Name + Environment.NewLine + "Training R²";
if (!dataTable.Rows.ContainsKey(trainingR2))
dataTable.Rows.Add(new DataRow(trainingR2));
var testR2 = result.Name + Environment.NewLine + " Test R²";
if (!dataTable.Rows.ContainsKey(testR2))
dataTable.Rows.Add(new DataRow(testR2));
dataTable.Rows[trainingR2].Values.Add(solution.TrainingRSquared);
dataTable.Rows[testR2].Values.Add(solution.TestRSquared);
}
return base.Apply();
}
public override IOperation Apply() {
ItemArray<StringValue> exceptions = ExceptionTreeParameter.ActualValue;
double count = exceptions.Count();
ResultCollection results = ResultsParameter.ActualValue;
DataTable exceptionFrequencies = ExceptionFrequenciesParameter.ActualValue;
if (exceptionFrequencies == null) {
exceptionFrequencies = new DataTable("Exception frequencies", "Relative frequency of exception aggregated over the whole population.");
exceptionFrequencies.VisualProperties.YAxisTitle = "Relative Exception Frequency";
ExceptionFrequenciesParameter.ActualValue = exceptionFrequencies;
results.Add(new Result("Exception frequencies", exceptionFrequencies));
}
// all rows must have the same number of values so we can just take the first
int numberOfValues = exceptionFrequencies.Rows.Select(r => r.Values.Count).DefaultIfEmpty().First();
foreach (var pair in exceptions.GroupBy(x => x.Value).ToDictionary(g => g.Key, g => (double)g.Count() / count)) {
string key = String.IsNullOrEmpty(pair.Key) ? "No Exception" : pair.Key;
if (!exceptionFrequencies.Rows.ContainsKey(key)) {
// initialize a new row for the symbol and pad with zeros
DataRow row = new DataRow(key, "", Enumerable.Repeat(0.0, numberOfValues));
row.VisualProperties.StartIndexZero = true;
exceptionFrequencies.Rows.Add(row);
}
exceptionFrequencies.Rows[key].Values.Add(Math.Round(pair.Value, 3));
}
// add a zero for each data row that was not modified in the previous loop
foreach (var row in exceptionFrequencies.Rows.Where(r => r.Values.Count != numberOfValues + 1))
row.Values.Add(0.0);
return base.Apply();
}
public override IOperation Apply() {
ItemArray<ISymbolicExpressionTree> expressions = SymbolicExpressionTreeParameter.ActualValue;
ResultCollection results = ResultCollection;
DataTable datatable;
if (VariableFrequenciesParameter.ActualValue == null) {
datatable = new DataTable("Variable frequencies", "Relative frequency of variable references aggregated over the whole population.");
datatable.VisualProperties.XAxisTitle = "Generation";
datatable.VisualProperties.YAxisTitle = "Relative Variable Frequency";
VariableFrequenciesParameter.ActualValue = datatable;
results.Add(new Result("Variable frequencies", "Relative frequency of variable references aggregated over the whole population.", datatable));
results.Add(new Result("Variable impacts", "The relative variable relevance calculated as the average relative variable frequency over the whole run.", new DoubleMatrix()));
}
datatable = VariableFrequenciesParameter.ActualValue;
// all rows must have the same number of values so we can just take the first
int numberOfValues = datatable.Rows.Select(r => r.Values.Count).DefaultIfEmpty().First();
foreach (var pair in SymbolicDataAnalysisVariableFrequencyAnalyzer.CalculateVariableFrequencies(expressions, AggregateLaggedVariables.Value)) {
if (!datatable.Rows.ContainsKey(pair.Key)) {
// initialize a new row for the variable and pad with zeros
DataRow row = new DataRow(pair.Key, "", Enumerable.Repeat(0.0, numberOfValues));
row.VisualProperties.StartIndexZero = true;
datatable.Rows.Add(row);
}
datatable.Rows[pair.Key].Values.Add(Math.Round(pair.Value, 3));
}
// add a zero for each data row that was not modified in the previous loop
foreach (var row in datatable.Rows.Where(r => r.Values.Count != numberOfValues + 1))
row.Values.Add(0.0);
// update variable impacts matrix
var orderedImpacts = (from row in datatable.Rows
select new { Name = row.Name, Impact = Math.Round(datatable.Rows[row.Name].Values.Average(), 3) })
.OrderByDescending(p => p.Impact)
.ToList();
var impacts = new DoubleMatrix();
var matrix = impacts as IStringConvertibleMatrix;
matrix.Rows = orderedImpacts.Count;
matrix.RowNames = orderedImpacts.Select(x => x.Name);
matrix.Columns = 1;
matrix.ColumnNames = new string[] { "Relative variable relevance" };
int i = 0;
foreach (var p in orderedImpacts) {
matrix.SetValue(p.Impact.ToString(), i++, 0);
}
VariableImpactsParameter.ActualValue = impacts;
results["Variable impacts"].Value = impacts;
return base.Apply();
}
private static IEnumerable<double> CalcCummulativeAverage(DataTable parent, DataTable other) {
IEnumerable<double> avg = Enumerable.Repeat(0.0, parent.Rows.First().Values.Count);
IEnumerable<double> real = Enumerable.Repeat(0.0, parent.Rows.First().Values.Count);
foreach (var rowName in parent.Rows.Select(x => x.Name)) {
var symboleParentRow = parent.Rows[rowName];
var symboleOtherRow = other.Rows[rowName];
avg = avg.Zip(symboleParentRow.Values, (x, y) => x + y);
real = real.Zip(symboleParentRow.Values.Zip(symboleOtherRow.Values, (x, y) => x * y), (x, y) => x + y);
}
return real.Zip(avg, (x, y) => x / y);
}
public override IOperation Apply() {
DataTable table = DataTableParameter.ActualValue;
if (table == null) {
table = new DataTable(DataTableParameter.ActualName);
DataTableParameter.ActualValue = table;
}
foreach (IParameter param in CollectedValues) {
ILookupParameter lookupParam = param as ILookupParameter;
string name = lookupParam != null ? lookupParam.TranslatedName : param.Name;
if (param.ActualValue is DoubleValue) {
AddValue(table, (param.ActualValue as DoubleValue).Value, name, param.Description);
} else if (param.ActualValue is IntValue) {
AddValue(table, (param.ActualValue as IntValue).Value, name, param.Description);
} else if (param.ActualValue is IEnumerable<DoubleValue>) {
IEnumerable<DoubleValue> values = (IEnumerable<DoubleValue>)param.ActualValue;
if (values.Count() <= 1) {
foreach (DoubleValue data in values)
AddValue(table, data != null ? data.Value : double.NaN, name, param.Description);
} else {
int counter = 1;
foreach (DoubleValue data in values) {
AddValue(table, data != null ? data.Value : double.NaN, name + " " + counter.ToString(), param.Description);
counter++;
}
}
} else if (param.ActualValue is IEnumerable<IntValue>) {
IEnumerable<IntValue> values = (IEnumerable<IntValue>)param.ActualValue;
if (values.Count() <= 1) {
foreach (IntValue data in values)
AddValue(table, data != null ? data.Value : double.NaN, name, param.Description);
} else {
int counter = 1;
foreach (IntValue data in values) {
AddValue(table, data != null ? data.Value : double.NaN, name + " " + counter.ToString(), param.Description);
counter++;
}
}
} else {
AddValue(table, double.NaN, name, param.Description);
}
}
return base.Apply();
}
protected override void Run(CancellationToken cancellationToken) {
// Set up the algorithm
if (SetSeedRandomly) Seed = new System.Random().Next();
var rand = new MersenneTwister((uint)Seed);
// Set up the results display
var iterations = new IntValue(0);
Results.Add(new Result("Iterations", iterations));
var table = new DataTable("Qualities");
table.Rows.Add(new DataRow("R² (train)"));
table.Rows.Add(new DataRow("R² (test)"));
Results.Add(new Result("Qualities", table));
var curLoss = new DoubleValue();
var curTestLoss = new DoubleValue();
Results.Add(new Result("R² (train)", curLoss));
Results.Add(new Result("R² (test)", curTestLoss));
var runCollection = new RunCollection();
if (StoreRuns)
Results.Add(new Result("Runs", runCollection));
// init
var problemData = Problem.ProblemData;
var targetVarName = problemData.TargetVariable;
var activeVariables = problemData.AllowedInputVariables.Concat(new string[] { problemData.TargetVariable });
var modifiableDataset = new ModifiableDataset(
activeVariables,
activeVariables.Select(v => problemData.Dataset.GetDoubleValues(v).ToList()));
var trainingRows = problemData.TrainingIndices;
var testRows = problemData.TestIndices;
var yPred = new double[trainingRows.Count()];
var yPredTest = new double[testRows.Count()];
var y = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TrainingIndices).ToArray();
var curY = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TrainingIndices).ToArray();
var yTest = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TestIndices).ToArray();
var curYTest = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TestIndices).ToArray();
var nu = Nu;
var mVars = (int)Math.Ceiling(M * problemData.AllowedInputVariables.Count());
var rRows = (int)Math.Ceiling(R * problemData.TrainingIndices.Count());
var alg = RegressionAlgorithm;
List<IRegressionModel> models = new List<IRegressionModel>();
try {
// Loop until iteration limit reached or canceled.
for (int i = 0; i < Iterations; i++) {
cancellationToken.ThrowIfCancellationRequested();
modifiableDataset.RemoveVariable(targetVarName);
modifiableDataset.AddVariable(targetVarName, curY.Concat(curYTest));
SampleTrainingData(rand, modifiableDataset, rRows, problemData.Dataset, curY, problemData.TargetVariable, problemData.TrainingIndices); // all training indices from the original problem data are allowed
var modifiableProblemData = new RegressionProblemData(modifiableDataset,
problemData.AllowedInputVariables.SampleRandomWithoutRepetition(rand, mVars),
problemData.TargetVariable);
modifiableProblemData.TrainingPartition.Start = 0;
modifiableProblemData.TrainingPartition.End = rRows;
modifiableProblemData.TestPartition.Start = problemData.TestPartition.Start;
modifiableProblemData.TestPartition.End = problemData.TestPartition.End;
if (!TrySetProblemData(alg, modifiableProblemData))
throw new NotSupportedException("The algorithm cannot be used with GBM.");
IRegressionModel model;
IRun run;
// try to find a model. The algorithm might fail to produce a model. In this case we just retry until the iterations are exhausted
if (TryExecute(alg, rand.Next(), RegressionAlgorithmResult, out model, out run)) {
int row = 0;
// update predictions for training and test
// update new targets (in the case of squared error loss we simply use negative residuals)
foreach (var pred in model.GetEstimatedValues(problemData.Dataset, trainingRows)) {
yPred[row] = yPred[row] + nu * pred;
curY[row] = y[row] - yPred[row];
row++;
}
row = 0;
foreach (var pred in model.GetEstimatedValues(problemData.Dataset, testRows)) {
yPredTest[row] = yPredTest[row] + nu * pred;
curYTest[row] = yTest[row] - yPredTest[row];
row++;
}
// determine quality
OnlineCalculatorError error;
var trainR = OnlinePearsonsRCalculator.Calculate(yPred, y, out error);
var testR = OnlinePearsonsRCalculator.Calculate(yPredTest, yTest, out error);
// iteration results
curLoss.Value = error == OnlineCalculatorError.None ? trainR * trainR : 0.0;
curTestLoss.Value = error == OnlineCalculatorError.None ? testR * testR : 0.0;
models.Add(model);
}
if (StoreRuns)
runCollection.Add(run);
table.Rows["R² (train)"].Values.Add(curLoss.Value);
table.Rows["R² (test)"].Values.Add(curTestLoss.Value);
iterations.Value = i + 1;
}
// produce solution
if (CreateSolution) {
// when all our models are symbolic models we can easily combine them to a single model
if (models.All(m => m is ISymbolicRegressionModel)) {
Results.Add(new Result("Solution", CreateSymbolicSolution(models, Nu, (IRegressionProblemData)problemData.Clone())));
}
// just produce an ensemble solution for now (TODO: correct scaling or linear regression for ensemble model weights)
var ensembleSolution = CreateEnsembleSolution(models, (IRegressionProblemData)problemData.Clone());
Results.Add(new Result("EnsembleSolution", ensembleSolution));
}
}
finally {
// reset everything
alg.Prepare(true);
}
}
protected DataTable(DataTable original, Cloner cloner)
: base(original, cloner) {
VisualProperties = (DataTableVisualProperties)cloner.Clone(original.visualProperties);
Rows = cloner.Clone(original.rows);
}
protected override void Run(CancellationToken cancellationToken) {
// Set up the algorithm
if (SetSeedRandomly) Seed = new System.Random().Next();
pyramid = new List<Population>();
seen.Clear();
random.Reset(Seed);
tracker = new EvaluationTracker(Problem, MaximumEvaluations);
// Set up the results display
Results.Add(new Result("Iterations", new IntValue(0)));
Results.Add(new Result("Evaluations", new IntValue(0)));
Results.Add(new Result("Best Solution", new BinaryVector(tracker.BestSolution)));
Results.Add(new Result("Best Quality", new DoubleValue(tracker.BestQuality)));
Results.Add(new Result("Evaluation Best Solution Was Found", new IntValue(tracker.BestFoundOnEvaluation)));
var table = new DataTable("Qualities");
table.Rows.Add(new DataRow("Best Quality"));
var iterationRows = new DataRow("Iteration Quality");
iterationRows.VisualProperties.LineStyle = DataRowVisualProperties.DataRowLineStyle.Dot;
table.Rows.Add(iterationRows);
Results.Add(new Result("Qualities", table));
table = new DataTable("Pyramid Levels");
table.Rows.Add(new DataRow("Levels"));
Results.Add(new Result("Pyramid Levels", table));
table = new DataTable("Stored Solutions");
table.Rows.Add(new DataRow("Solutions"));
Results.Add(new Result("Stored Solutions", table));
// Loop until iteration limit reached or canceled.
for (ResultsIterations = 0; ResultsIterations < MaximumIterations; ResultsIterations++) {
double fitness = double.NaN;
try {
fitness = iterate();
cancellationToken.ThrowIfCancellationRequested();
} finally {
ResultsEvaluations = tracker.Evaluations;
ResultsBestSolution = new BinaryVector(tracker.BestSolution);
ResultsBestQuality = tracker.BestQuality;
ResultsBestFoundOnEvaluation = tracker.BestFoundOnEvaluation;
ResultsQualitiesBest.Values.Add(tracker.BestQuality);
ResultsQualitiesIteration.Values.Add(fitness);
ResultsLevels.Values.Add(pyramid.Count);
ResultsSolutions.Values.Add(seen.Count);
}
}
}
protected override void Run(CancellationToken cancellationToken) {
// Set up the algorithm
if (SetSeedRandomly) Seed = new System.Random().Next();
// Set up the results display
var iterations = new IntValue(0);
Results.Add(new Result("Iterations", iterations));
var table = new DataTable("Qualities");
table.Rows.Add(new DataRow("Loss (train)"));
table.Rows.Add(new DataRow("Loss (test)"));
Results.Add(new Result("Qualities", table));
var curLoss = new DoubleValue();
Results.Add(new Result("Loss (train)", curLoss));
// init
var problemData = (IRegressionProblemData)Problem.ProblemData.Clone();
var lossFunction = LossFunctionParameter.Value;
var state = GradientBoostedTreesAlgorithmStatic.CreateGbmState(problemData, lossFunction, (uint)Seed, MaxSize, R, M, Nu);
var updateInterval = UpdateIntervalParameter.Value.Value;
// Loop until iteration limit reached or canceled.
for (int i = 0; i < Iterations; i++) {
cancellationToken.ThrowIfCancellationRequested();
GradientBoostedTreesAlgorithmStatic.MakeStep(state);
// iteration results
if (i % updateInterval == 0) {
curLoss.Value = state.GetTrainLoss();
table.Rows["Loss (train)"].Values.Add(curLoss.Value);
table.Rows["Loss (test)"].Values.Add(state.GetTestLoss());
iterations.Value = i;
}
}
// final results
iterations.Value = Iterations;
curLoss.Value = state.GetTrainLoss();
table.Rows["Loss (train)"].Values.Add(curLoss.Value);
table.Rows["Loss (test)"].Values.Add(state.GetTestLoss());
// produce variable relevance
var orderedImpacts = state.GetVariableRelevance().Select(t => new { name = t.Key, impact = t.Value }).ToList();
var impacts = new DoubleMatrix();
var matrix = impacts as IStringConvertibleMatrix;
matrix.Rows = orderedImpacts.Count;
matrix.RowNames = orderedImpacts.Select(x => x.name);
matrix.Columns = 1;
matrix.ColumnNames = new string[] { "Relative variable relevance" };
int rowIdx = 0;
foreach (var p in orderedImpacts) {
matrix.SetValue(string.Format("{0:N2}", p.impact), rowIdx++, 0);
}
Results.Add(new Result("Variable relevance", impacts));
Results.Add(new Result("Loss (test)", new DoubleValue(state.GetTestLoss())));
// produce solution
if (CreateSolution) {
var model = state.GetModel();
// for logistic regression we produce a classification solution
if (lossFunction is LogisticRegressionLoss) {
var classificationModel = new DiscriminantFunctionClassificationModel(model,
new AccuracyMaximizationThresholdCalculator());
var classificationProblemData = new ClassificationProblemData(problemData.Dataset,
problemData.AllowedInputVariables, problemData.TargetVariable, problemData.Transformations);
classificationModel.RecalculateModelParameters(classificationProblemData, classificationProblemData.TrainingIndices);
var classificationSolution = new DiscriminantFunctionClassificationSolution(classificationModel, classificationProblemData);
Results.Add(new Result("Solution", classificationSolution));
} else {
// otherwise we produce a regression solution
Results.Add(new Result("Solution", new RegressionSolution(model, problemData)));
}
}
}
public override IOperation Apply() {
int updateInterval = UpdateIntervalParameter.Value.Value;
IntValue updateCounter = UpdateCounterParameter.ActualValue;
// if counter does not yet exist then initialize it with update interval
// to make sure the solutions are analyzed on the first application of this operator
if (updateCounter == null) {
updateCounter = new IntValue(updateInterval);
UpdateCounterParameter.ActualValue = updateCounter;
} else updateCounter.Value++;
//analyze solutions only every 'updateInterval' times
if (updateCounter.Value == updateInterval) {
updateCounter.Value = 0;
bool max = MaximizationParameter.ActualValue.Value;
ItemArray<DoubleValue> qualities = QualityParameter.ActualValue;
bool storeHistory = StoreHistoryParameter.Value.Value;
int count = CurrentScopeParameter.ActualValue.SubScopes.Count;
if (count > 1) {
// calculate solution similarities
var similarityMatrix = SimilarityCalculator.CalculateSolutionCrowdSimilarity(CurrentScopeParameter.ActualValue);
// sort similarities by quality
double[][] sortedSimilarityMatrix = null;
if (max)
sortedSimilarityMatrix = similarityMatrix
.Select((x, index) => new { Solutions = x, Quality = qualities[index] })
.OrderByDescending(x => x.Quality)
.Select(x => x.Solutions)
.ToArray();
else
sortedSimilarityMatrix = similarityMatrix
.Select((x, index) => new { Solutions = x, Quality = qualities[index] })
.OrderBy(x => x.Quality)
.Select(x => x.Solutions)
.ToArray();
double[,] similarities = new double[similarityMatrix.Length, similarityMatrix[0].Length];
for (int i = 0; i < similarityMatrix.Length; i++)
for (int j = 0; j < similarityMatrix[0].Length; j++)
similarities[i, j] = similarityMatrix[i][j];
// calculate minimum, average and maximum similarities
double similarity;
double[] minSimilarities = new double[count];
double[] avgSimilarities = new double[count];
double[] maxSimilarities = new double[count];
for (int i = 0; i < count; i++) {
minSimilarities[i] = 1;
avgSimilarities[i] = 0;
maxSimilarities[i] = 0;
for (int j = 0; j < count; j++) {
if (i != j) {
similarity = similarities[i, j];
if ((similarity < 0) || (similarity > 1))
throw new InvalidOperationException("Solution similarities have to be in the interval [0;1].");
if (minSimilarities[i] > similarity) minSimilarities[i] = similarity;
avgSimilarities[i] += similarity;
if (maxSimilarities[i] < similarity) maxSimilarities[i] = similarity;
}
}
avgSimilarities[i] = avgSimilarities[i] / (count - 1);
}
double avgMinSimilarity = minSimilarities.Average();
double avgAvgSimilarity = avgSimilarities.Average();
double avgMaxSimilarity = maxSimilarities.Average();
// fetch results collection
ResultCollection results;
if (!ResultsParameter.ActualValue.ContainsKey(Name + " Results")) {
results = new ResultCollection();
ResultsParameter.ActualValue.Add(new Result(Name + " Results", results));
} else {
results = (ResultCollection)ResultsParameter.ActualValue[Name + " Results"].Value;
}
// store similarities
HeatMap similaritiesHeatMap = new HeatMap(similarities, "Solution Similarities", 0.0, 1.0);
if (!results.ContainsKey("Solution Similarities"))
results.Add(new Result("Solution Similarities", similaritiesHeatMap));
else
results["Solution Similarities"].Value = similaritiesHeatMap;
// store similarities history
if (storeHistory) {
if (!results.ContainsKey("Solution Similarities History")) {
HeatMapHistory history = new HeatMapHistory();
history.Add(similaritiesHeatMap);
results.Add(new Result("Solution Similarities History", history));
} else {
((HeatMapHistory)results["Solution Similarities History"].Value).Add(similaritiesHeatMap);
}
}
// store average minimum, average and maximum similarity
if (!results.ContainsKey("Average Minimum Solution Similarity"))
results.Add(new Result("Average Minimum Solution Similarity", new DoubleValue(avgMinSimilarity)));
else
((DoubleValue)results["Average Minimum Solution Similarity"].Value).Value = avgMinSimilarity;
if (!results.ContainsKey("Average Average Solution Similarity"))
results.Add(new Result("Average Average Solution Similarity", new DoubleValue(avgAvgSimilarity)));
else
((DoubleValue)results["Average Average Solution Similarity"].Value).Value = avgAvgSimilarity;
if (!results.ContainsKey("Average Maximum Solution Similarity"))
results.Add(new Result("Average Maximum Solution Similarity", new DoubleValue(avgMaxSimilarity)));
else
((DoubleValue)results["Average Maximum Solution Similarity"].Value).Value = avgMaxSimilarity;
// store average minimum, average and maximum solution similarity data table
DataTable minAvgMaxSimilarityDataTable;
if (!results.ContainsKey("Average Minimum/Average/Maximum Solution Similarity")) {
minAvgMaxSimilarityDataTable = new DataTable("Average Minimum/Average/Maximum Solution Similarity");
minAvgMaxSimilarityDataTable.VisualProperties.XAxisTitle = "Iteration";
minAvgMaxSimilarityDataTable.VisualProperties.YAxisTitle = "Solution Similarity";
minAvgMaxSimilarityDataTable.Rows.Add(new DataRow("Average Minimum Solution Similarity", null));
minAvgMaxSimilarityDataTable.Rows["Average Minimum Solution Similarity"].VisualProperties.StartIndexZero = true;
minAvgMaxSimilarityDataTable.Rows.Add(new DataRow("Average Average Solution Similarity", null));
minAvgMaxSimilarityDataTable.Rows["Average Average Solution Similarity"].VisualProperties.StartIndexZero = true;
minAvgMaxSimilarityDataTable.Rows.Add(new DataRow("Average Maximum Solution Similarity", null));
minAvgMaxSimilarityDataTable.Rows["Average Maximum Solution Similarity"].VisualProperties.StartIndexZero = true;
results.Add(new Result("Average Minimum/Average/Maximum Solution Similarity", minAvgMaxSimilarityDataTable));
} else {
minAvgMaxSimilarityDataTable = (DataTable)results["Average Minimum/Average/Maximum Solution Similarity"].Value;
}
minAvgMaxSimilarityDataTable.Rows["Average Minimum Solution Similarity"].Values.Add(avgMinSimilarity);
minAvgMaxSimilarityDataTable.Rows["Average Average Solution Similarity"].Values.Add(avgAvgSimilarity);
minAvgMaxSimilarityDataTable.Rows["Average Maximum Solution Similarity"].Values.Add(avgMaxSimilarity);
// store minimum, average, maximum similarities data table
DataTable minAvgMaxSimilaritiesDataTable = new DataTable("Minimum/Average/Maximum Solution Similarities");
minAvgMaxSimilaritiesDataTable.VisualProperties.XAxisTitle = "Solution Index";
minAvgMaxSimilaritiesDataTable.VisualProperties.YAxisTitle = "Solution Similarity";
minAvgMaxSimilaritiesDataTable.Rows.Add(new DataRow("Minimum Solution Similarity", null, minSimilarities));
minAvgMaxSimilaritiesDataTable.Rows["Minimum Solution Similarity"].VisualProperties.ChartType = DataRowVisualProperties.DataRowChartType.Points;
minAvgMaxSimilaritiesDataTable.Rows.Add(new DataRow("Average Solution Similarity", null, avgSimilarities));
minAvgMaxSimilaritiesDataTable.Rows["Average Solution Similarity"].VisualProperties.ChartType = DataRowVisualProperties.DataRowChartType.Points;
minAvgMaxSimilaritiesDataTable.Rows.Add(new DataRow("Maximum Solution Similarity", null, maxSimilarities));
minAvgMaxSimilaritiesDataTable.Rows["Maximum Solution Similarity"].VisualProperties.ChartType = DataRowVisualProperties.DataRowChartType.Points;
if (!results.ContainsKey("Minimum/Average/Maximum Solution Similarities")) {
results.Add(new Result("Minimum/Average/Maximum Solution Similarities", minAvgMaxSimilaritiesDataTable));
} else {
results["Minimum/Average/Maximum Solution Similarities"].Value = minAvgMaxSimilaritiesDataTable;
}
// store minimum, average, maximum similarities history
if (storeHistory) {
if (!results.ContainsKey("Minimum/Average/Maximum Solution Similarities History")) {
DataTableHistory history = new DataTableHistory();
history.Add(minAvgMaxSimilaritiesDataTable);
results.Add(new Result("Minimum/Average/Maximum Solution Similarities History", history));
} else {
((DataTableHistory)results["Minimum/Average/Maximum Solution Similarities History"].Value).Add(minAvgMaxSimilaritiesDataTable);
}
}
}
}
return base.Apply();
}
private void AddAverageTableEntry(IntValue[] values, string tableName) {
if (!ResultCollection.ContainsKey(tableName)) {
var newTable = new DataTable(tableName, "");
newTable.VisualProperties.YAxisTitle = "Average";
newTable.VisualProperties.YAxisMaximumAuto = false;
DataRow row = new DataRow("Average");
row.VisualProperties.StartIndexZero = true;
newTable.Rows.Add(row);
ResultCollection.Add(new Result(tableName, newTable));
}
var table = ((DataTable)ResultCollection[tableName].Value);
table.Rows["Average"].Values.Add(values.Select(x => x.Value).Average());
}
public override IOperation Apply() {
ItemArray<ISymbolicExpressionTree> expressions = SymbolicExpressionTreeParameter.ActualValue;
ResultCollection results = ResultsParameter.ActualValue;
DataTable symbolFrequencies = SymbolFrequenciesParameter.ActualValue;
if (symbolFrequencies == null) {
symbolFrequencies = new DataTable("Symbol frequencies", "Relative frequency of symbols aggregated over the whole population.");
symbolFrequencies.VisualProperties.YAxisTitle = "Relative Symbol Frequency";
SymbolFrequenciesParameter.ActualValue = symbolFrequencies;
results.Add(new Result("Symbol frequencies", symbolFrequencies));
}
// all rows must have the same number of values so we can just take the first
int numberOfValues = symbolFrequencies.Rows.Select(r => r.Values.Count).DefaultIfEmpty().First();
foreach (var pair in SymbolicExpressionSymbolFrequencyAnalyzer.CalculateSymbolFrequencies(expressions, AggregrateSymbolsWithDifferentSubtreeCount.Value)) {
if (!symbolFrequencies.Rows.ContainsKey(pair.Key)) {
// initialize a new row for the symbol and pad with zeros
DataRow row = new DataRow(pair.Key, "", Enumerable.Repeat(0.0, numberOfValues));
row.VisualProperties.StartIndexZero = true;
symbolFrequencies.Rows.Add(row);
}
symbolFrequencies.Rows[pair.Key].Values.Add(Math.Round(pair.Value, 3));
}
// add a zero for each data row that was not modified in the previous loop
foreach (var row in symbolFrequencies.Rows.Where(r => r.Values.Count != numberOfValues + 1))
row.Values.Add(0.0);
return base.Apply();
}
public override IOperation Apply() {
//code executed in the first call of analyzer
if (qualitiesBeforeCoOp == null) {
double[] trainingQuality;
// sort is ascending and we take the first n% => order so that best solutions are smallest
// sort order is determined by maximization parameter
if (Maximization.Value) {
// largest values must be sorted first
trainingQuality = Quality.Select(x => -x.Value).ToArray();
} else {
// smallest values must be sorted first
trainingQuality = Quality.Select(x => x.Value).ToArray();
}
// sort trees by training qualities
int topN = (int)Math.Max(trainingQuality.Length * PercentageOfBestSolutions, 1);
scopeIndexes = Enumerable.Range(0, trainingQuality.Length).ToArray();
Array.Sort(trainingQuality, scopeIndexes);
scopeIndexes = scopeIndexes.Take(topN).ToArray();
qualitiesBeforeCoOp = scopeIndexes.Select(x => Quality[x].Value).ToArray();
OperationCollection operationCollection = new OperationCollection();
operationCollection.Parallel = true;
foreach (var scopeIndex in scopeIndexes) {
var childOperation = ExecutionContext.CreateChildOperation(ConstantOptimizationEvaluator, ExecutionContext.Scope.SubScopes[scopeIndex]);
operationCollection.Add(childOperation);
}
return new OperationCollection { operationCollection, ExecutionContext.CreateOperation(this) };
}
//code executed to analyze results of constant optimization
double[] qualitiesAfterCoOp = scopeIndexes.Select(x => Quality[x].Value).ToArray();
var qualityImprovement = qualitiesBeforeCoOp.Zip(qualitiesAfterCoOp, (b, a) => a - b).ToArray();
if (!ResultCollection.ContainsKey(DataTableNameConstantOptimizationImprovement)) {
var dataTable = new DataTable(DataTableNameConstantOptimizationImprovement);
ResultCollection.Add(new Result(DataTableNameConstantOptimizationImprovement, dataTable));
dataTable.VisualProperties.YAxisTitle = "R²";
dataTable.Rows.Add(new DataRow(DataRowNameMinimumImprovement));
MinimumImprovement.VisualProperties.StartIndexZero = true;
dataTable.Rows.Add(new DataRow(DataRowNameMedianImprovement));
MedianImprovement.VisualProperties.StartIndexZero = true;
dataTable.Rows.Add(new DataRow(DataRowNameAverageImprovement));
AverageImprovement.VisualProperties.StartIndexZero = true;
dataTable.Rows.Add(new DataRow(DataRowNameMaximumImprovement));
MaximumImprovement.VisualProperties.StartIndexZero = true;
}
MinimumImprovement.Values.Add(qualityImprovement.Min());
MedianImprovement.Values.Add(qualityImprovement.Median());
AverageImprovement.Values.Add(qualityImprovement.Average());
MaximumImprovement.Values.Add(qualityImprovement.Max());
qualitiesBeforeCoOp = null;
scopeIndexes = null;
return base.Apply();
}
private void AddSemanticallyDifferentFromRootedParentTableEntry(BoolValue[] semanticallyDifferentFromRootedParent) {
if (SemanticallyDifferentFromRootedParentDataTable == null) {
var table = new DataTable(SemanticallyDifferentFromRootedParentParameterName, "");
table.VisualProperties.YAxisTitle = "Percentage";
table.VisualProperties.YAxisMaximumFixedValue = 100.0;
table.VisualProperties.YAxisMaximumAuto = false;
DataRow differentRow = new DataRow("Different From Parent");
differentRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(differentRow);
DataRow sameRow = new DataRow("Same As Parent");
sameRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(sameRow);
SemanticallyDifferentFromRootedParentDataTable = table;
}
double different = semanticallyDifferentFromRootedParent.Count(x => x.Value);
SemanticallyDifferentFromRootedParentDataTable.Rows["Different From Parent"].Values.Add(different / semanticallyDifferentFromRootedParent.Length * 100.0);
SemanticallyDifferentFromRootedParentDataTable.Rows["Same As Parent"].Values.Add((semanticallyDifferentFromRootedParent.Length - different) / semanticallyDifferentFromRootedParent.Length * 100.0);
}
public override IOperation Apply() {
ResultCollection results = ResultsParameter.ActualValue;
List<IScope> scopes = new List<IScope>() { ExecutionContext.Scope };
for (int i = 0; i < DepthParameter.Value.Value; i++)
scopes = scopes.Select(x => (IEnumerable<IScope>)x.SubScopes).Aggregate((a, b) => a.Concat(b)).ToList();
ItemCollection<StringValue> collectedValues = CollectedValuesParameter.Value;
foreach (StringValue collected in collectedValues) {
//collect the values of the successful offspring
Dictionary<String, int> counts = new Dictionary<String, int>();
for (int i = 0; i < scopes.Count; i++) {
IScope child = scopes[i];
string successfulOffspringFlag = SuccessfulOffspringFlagParameter.Value.Value;
if (child.Variables.ContainsKey(collected.Value) &&
child.Variables.ContainsKey(successfulOffspringFlag) &&
(child.Variables[successfulOffspringFlag].Value is BoolValue) &&
(child.Variables[successfulOffspringFlag].Value as BoolValue).Value) {
String key = child.Variables[collected.Value].Value.ToString();
if (!counts.ContainsKey(key))
counts.Add(key, 1);
else
counts[key]++;
}
}
if (counts.Count > 0) {
//create a data table containing the collected values
ResultCollection successfulOffspringAnalysis;
if (SuccessfulOffspringAnalysisParameter.ActualValue == null) {
successfulOffspringAnalysis = new ResultCollection();
SuccessfulOffspringAnalysisParameter.ActualValue = successfulOffspringAnalysis;
} else {
successfulOffspringAnalysis = SuccessfulOffspringAnalysisParameter.ActualValue;
}
string resultKey = "SuccessfulOffspringAnalyzer Results";
if (!results.ContainsKey(resultKey)) {
results.Add(new Result(resultKey, successfulOffspringAnalysis));
} else {
results[resultKey].Value = successfulOffspringAnalysis;
}
DataTable successProgressAnalysis;
if (!successfulOffspringAnalysis.ContainsKey(collected.Value)) {
successProgressAnalysis = new DataTable();
successProgressAnalysis.Name = collected.Value;
successfulOffspringAnalysis.Add(new Result(collected.Value, successProgressAnalysis));
} else {
successProgressAnalysis = successfulOffspringAnalysis[collected.Value].Value as DataTable;
}
int successfulCount = 0;
foreach (string key in counts.Keys) {
successfulCount += counts[key];
}
foreach (String value in counts.Keys) {
DataRow row;
if (!successProgressAnalysis.Rows.ContainsKey(value)) {
row = new DataRow(value);
int iterations = GenerationsParameter.ActualValue.Value;
//fill up all values seen the first time
for (int i = 1; i < iterations; i++)
row.Values.Add(0);
successProgressAnalysis.Rows.Add(row);
} else {
row = successProgressAnalysis.Rows[value];
}
row.Values.Add(counts[value] / (double)successfulCount);
}
//fill up all values that are not present in the current generation
foreach (DataRow row in successProgressAnalysis.Rows) {
if (!counts.ContainsKey(row.Name))
row.Values.Add(0);
}
}
}
return base.Apply();
}
private void AddValue(DataTable table, double data, string name, string description) {
DataRow row;
table.Rows.TryGetValue(name, out row);
if (row == null) {
row = new DataRow(name, description);
row.VisualProperties.StartIndexZero = StartIndexZero;
row.Values.Add(data);
table.Rows.Add(row);
} else {
row.Values.Add(data);
}
}
private void AddSemanticLocalityTableEntry(double average) {
if (SemanticLocalityDataTable == null) {
var table = new DataTable(SemanticLocalityParameterName, "");
table.VisualProperties.YAxisTitle = "Average Fitness Change";
DataRow row = new DataRow("Sematic Locality");
row.VisualProperties.StartIndexZero = true;
table.Rows.Add(row);
SemanticLocalityDataTable = table;
}
SemanticLocalityDataTable.Rows["Sematic Locality"].Values.Add(average);
}
public override IOperation Apply() {
DataTable ageDistribution = null;
ResultCollection results = ResultsParameter.ActualValue;
string description = "Shows the age distributions in the current population.";
if (results.ContainsKey(HistogramName)) {
ageDistribution = results[HistogramName].Value as DataTable;
} else {
ageDistribution = new DataTable("Population Age Distribution", description);
ageDistribution.VisualProperties.XAxisTitle = AgeParameter.ActualName;
ageDistribution.VisualProperties.YAxisTitle = "Frequency";
results.Add(new Result(HistogramName, description, ageDistribution));
}
DataRow row;
if (!ageDistribution.Rows.TryGetValue("AgeDistribution", out row)) {
row = new DataRow("AgeDistribution");
row.VisualProperties.ChartType = DataRowVisualProperties.DataRowChartType.Histogram;
ageDistribution.Rows.Add(row);
}
var ages = AgeParameter.ActualValue;
row.Values.Replace(ages.Select(x => x.Value));
if (StoreHistory) {
string historyResultName = HistogramName + " History";
DataTableHistory adHistory = null;
if (results.ContainsKey(historyResultName)) {
adHistory = results[historyResultName].Value as DataTableHistory;
} else {
adHistory = new DataTableHistory();
results.Add(new Result(historyResultName, adHistory));
}
DataTable table = (DataTable)ageDistribution.Clone();
IntValue iteration = IterationsParameter.ActualValue;
if (iteration != null) {
string iterationName = IterationsParameter.ActualName;
if (iterationName.EndsWith("s")) iterationName = iterationName.Remove(iterationName.Length - 1);
string appendix = " at " + iterationName + " " + iteration.Value.ToString();
table.Name += appendix;
table.Rows["AgeDistribution"].VisualProperties.DisplayName += appendix;
}
adHistory.Add(table);
}
return base.Apply();
}
public override IOperation Apply() {
var results = ResultCollectionParameter.ActualValue;
if (!results.ContainsKey(RegressionSolutionQualitiesResultName)) {
var newDataTable = new DataTable(RegressionSolutionQualitiesResultName);
results.Add(new Result(RegressionSolutionQualitiesResultName, "Chart displaying the training and test qualities of the regression solutions.", newDataTable));
}
var dataTable = (DataTable)results[RegressionSolutionQualitiesResultName].Value;
// only if the parameters are available (not available in old persisted code)
ILookupParameter<DoubleValue> trainingQualityParam = null;
ILookupParameter<DoubleValue> testQualityParam = null;
// store actual names of parameter because it is changed below
trainingQualityParam = TrainingQualityParameter;
string prevTrainingQualityParamName = trainingQualityParam.ActualName;
testQualityParam = TestQualityParameter;
string prevTestQualityParamName = testQualityParam.ActualName;
foreach (var result in results.Where(r => r.Value is IRegressionSolution)) {
var solution = (IRegressionSolution)result.Value;
var trainingR2Name = result.Name + " Training R²";
if (!dataTable.Rows.ContainsKey(trainingR2Name))
dataTable.Rows.Add(new DataRow(trainingR2Name));
var testR2Name = result.Name + " Test R²";
if (!dataTable.Rows.ContainsKey(testR2Name))
dataTable.Rows.Add(new DataRow(testR2Name));
dataTable.Rows[trainingR2Name].Values.Add(solution.TrainingRSquared);
dataTable.Rows[testR2Name].Values.Add(solution.TestRSquared);
// also add training and test R² to the scope using the parameters
// HACK: we change the ActualName of the parameter to write two variables for each solution in the results collection
trainingQualityParam.ActualName = trainingR2Name;
trainingQualityParam.ActualValue = new DoubleValue(solution.TrainingRSquared);
testQualityParam.ActualName = testR2Name;
testQualityParam.ActualValue = new DoubleValue(solution.TestRSquared);
}
trainingQualityParam.ActualName = prevTrainingQualityParamName;
testQualityParam.ActualName = prevTestQualityParamName;
return base.Apply();
}
private static DataTable CalcAverage(RunCollection runs, string tableName) {
DataTable temptable = new DataTable();
var visibleRuns = runs.Where(r => r.Visible);
var resultName = (string)tableName;
var dataTables = visibleRuns.Where(r => r.Results.ContainsKey(resultName)).Select(r => (DataTable)r.Results[resultName]);
if (dataTables.Count() != visibleRuns.Count()) {
throw new ArgumentException("Should not happen");
}
var dataRows = dataTables.SelectMany(dt => dt.Rows).GroupBy(r => r.Name, r => r);
foreach (var row in dataRows) {
var averageValues = DataRowsAggregate(Enumerable.Average, row.Select(r => r.Values));
DataRow averageRow = new DataRow(row.Key, "Average of Values", averageValues);
temptable.Rows.Add(averageRow);
}
return temptable;
}
private void AddConstructiveEffectTableEntry(IntValue[] constructiveEffect) {
if (ConstructiveEffectDataTable == null) {
var table = new DataTable(ConstructiveEffectParameterName, "");
table.VisualProperties.YAxisTitle = "Percentage";
table.VisualProperties.YAxisMaximumFixedValue = 100.0;
table.VisualProperties.YAxisMaximumAuto = false;
DataRow worseThanRootedRow = new DataRow("Worse than rooted");
worseThanRootedRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(worseThanRootedRow);
DataRow betterThanRootedRow = new DataRow("Better than rooted");
betterThanRootedRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(betterThanRootedRow);
DataRow betterThanBothRow = new DataRow("Better than both");
betterThanBothRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(betterThanBothRow);
ConstructiveEffectDataTable = table;
}
List<double> constructiveEffectCount = new List<double>() { 0.0, 0.0, 0.0 };
for (int i = 0; i < constructiveEffect.Length; i++) {
constructiveEffectCount[constructiveEffect[i].Value]++;
}
ConstructiveEffectDataTable.Rows["Worse than rooted"].Values.Add(constructiveEffectCount[0] / constructiveEffect.Length * 100.0);
ConstructiveEffectDataTable.Rows["Better than rooted"].Values.Add((constructiveEffectCount[1] + constructiveEffectCount[2]) / constructiveEffect.Length * 100.0);
ConstructiveEffectDataTable.Rows["Better than both"].Values.Add(constructiveEffectCount[2] / constructiveEffect.Length * 100.0);
}
public override IOperation Apply() {
ItemArray<BoolArray> cases = CasesParameter.ActualValue;
ItemArray<DoubleArray> caseQualities = CaseQualitiesParameter.ActualValue;
int length = cases.First().Length;
if (cases.Any(x => x.Length != length) || caseQualities.Any(x => x.Length != length)) {
throw new ArgumentException("Every individual has to have the same number of cases.");
}
double count = cases.Count();
int rows = cases.First().Length;
ResultCollection results = ResultsParameter.ActualValue;
DataTable casesSolvedFrequencies = CasesSolvedFrequenciesParameter.ActualValue;
DataTable summedCaseQualities = SummedCaseQualitiesParameter.ActualValue;
if (casesSolvedFrequencies == null || summedCaseQualities == null) {
casesSolvedFrequencies = new DataTable("Cases solved", "Absolute frequency of cases solved aggregated over the whole population.");
casesSolvedFrequencies.VisualProperties.YAxisTitle = "Absolute Cases Solved Frequency";
summedCaseQualities = new DataTable("Summed case qualities", "Absolute frequency of cases solved aggregated over the whole population.");
summedCaseQualities.VisualProperties.YAxisTitle = "Summed Cases Qualities";
CasesSolvedFrequenciesParameter.ActualValue = casesSolvedFrequencies;
SummedCaseQualitiesParameter.ActualValue = summedCaseQualities;
results.Add(new Result("Cases solved", casesSolvedFrequencies));
results.Add(new Result("Summed cases qualities", summedCaseQualities));
}
// all rows must have the same number of values so we can just take the first
int numberOfValues = casesSolvedFrequencies.Rows.Select(r => r.Values.Count).DefaultIfEmpty().First();
foreach (var triple in Enumerable.Range(0, cases[0].Length).Select(i => new Tuple<string, int, double>("Case " + i, cases.Count(caseArray => caseArray[i]), caseQualities.Sum(casesQualityArray => casesQualityArray[i])))) {
if (!casesSolvedFrequencies.Rows.ContainsKey(triple.Item1)) {
// initialize a new row for the symbol and pad with zeros
DataRow row = new DataRow(triple.Item1, "", Enumerable.Repeat(0.0, numberOfValues));
row.VisualProperties.StartIndexZero = true;
casesSolvedFrequencies.Rows.Add(row);
row = new DataRow(triple.Item1, "", Enumerable.Repeat(0.0, numberOfValues));
row.VisualProperties.StartIndexZero = true;
summedCaseQualities.Rows.Add(row);
}
casesSolvedFrequencies.Rows[triple.Item1].Values.Add(triple.Item2);
summedCaseQualities.Rows[triple.Item1].Values.Add(triple.Item3);
}
return base.Apply();
}
public RunCollectionChartAverageView() {
InitializeComponent();
combinedDataTable = new DataTable("Combined DataTable", "A data table containing data rows from multiple runs.");
viewHost.Content = combinedDataTable;
suppressUpdates = false;
}
public override IOperation Apply() {
var sp = StrategyParametersParameter.ActualValue;
var vector = MeanParameter.ActualValue;
var results = ResultsParameter.ActualValue;
var qualities = QualityParameter.ActualValue;
double min = qualities[0].Value, max = qualities[0].Value, avg = qualities[0].Value;
for (int i = 1; i < qualities.Length; i++) {
if (qualities[i].Value < min) min = qualities[i].Value;
if (qualities[i].Value > max) max = qualities[i].Value;
avg += qualities[i].Value;
}
avg /= qualities.Length;
DataTable progress;
if (results.ContainsKey("Progress")) {
progress = (DataTable)results["Progress"].Value;
} else {
progress = new DataTable("Progress");
progress.Rows.Add(new DataRow("AxisRatio"));
progress.Rows.Add(new DataRow("Sigma"));
progress.Rows.Add(new DataRow("Min Quality"));
progress.Rows.Add(new DataRow("Max Quality"));
progress.Rows.Add(new DataRow("Avg Quality"));
progress.VisualProperties.YAxisLogScale = true;
results.Add(new Result("Progress", progress));
}
progress.Rows["AxisRatio"].Values.Add(sp.AxisRatio);
progress.Rows["Sigma"].Values.Add(sp.Sigma);
progress.Rows["Min Quality"].Values.Add(min);
progress.Rows["Max Quality"].Values.Add(max);
progress.Rows["Avg Quality"].Values.Add(avg);
DataTable scaling;
if (results.ContainsKey("Scaling")) {
scaling = (DataTable)results["Scaling"].Value;
} else {
scaling = new DataTable("Scaling");
scaling.VisualProperties.YAxisLogScale = true;
for (int i = 0; i < sp.C.GetLength(0); i++)
scaling.Rows.Add(new DataRow("Axis" + i.ToString()));
results.Add(new Result("Scaling", scaling));
}
for (int i = 0; i < sp.C.GetLength(0); i++)
scaling.Rows["Axis" + i.ToString()].Values.Add(sp.D[i]);
DataTable realVector;
if (results.ContainsKey("Object Variables")) {
realVector = (DataTable)results["Object Variables"].Value;
} else {
realVector = new DataTable("Object Variables");
for (int i = 0; i < vector.Length; i++)
realVector.Rows.Add(new DataRow("Axis" + i.ToString()));
results.Add(new Result("Object Variables", realVector));
}
for (int i = 0; i < vector.Length; i++)
realVector.Rows["Axis" + i.ToString()].Values.Add(vector[i]);
DataTable stdDevs;
if (results.ContainsKey("Standard Deviations")) {
stdDevs = (DataTable)results["Standard Deviations"].Value;
} else {
stdDevs = new DataTable("Standard Deviations");
stdDevs.VisualProperties.YAxisLogScale = true;
stdDevs.Rows.Add(new DataRow("MinStdDev"));
stdDevs.Rows.Add(new DataRow("MaxStdDev"));
for (int i = 0; i < vector.Length; i++)
stdDevs.Rows.Add(new DataRow("Axis" + i.ToString()));
results.Add(new Result("Standard Deviations", stdDevs));
}
for (int i = 0; i < vector.Length; i++)
stdDevs.Rows["Axis" + i.ToString()].Values.Add(Math.Sqrt(sp.C[i, i]));
stdDevs.Rows["MinStdDev"].Values.Add(sp.D.Min() * sp.Sigma);
stdDevs.Rows["MaxStdDev"].Values.Add(sp.D.Max() * sp.Sigma);
return base.Apply();
}
private void AddSemanticallyEquivalentCrossoverTableEntry(IntValue[] semanticallyEquivalentCrossover) {
if (SemanticallyEquivalentCrossoverDataTable == null) {
var table = new DataTable(SemanticallyEquivalentCrossoverParameterName, "");
table.VisualProperties.YAxisTitle = "Percentage";
table.VisualProperties.YAxisMaximumFixedValue = 100.0;
table.VisualProperties.YAxisMaximumAuto = false;
DataRow noCrossoverRow = new DataRow("No Crossover");
noCrossoverRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(noCrossoverRow);
DataRow equivalentRow = new DataRow("Equivalent");
equivalentRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(equivalentRow);
DataRow equivalentOrNoCrossoverRow = new DataRow("Equivalent + No Crossover");
equivalentOrNoCrossoverRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(equivalentOrNoCrossoverRow);
DataRow differentRow = new DataRow("Different");
differentRow.VisualProperties.StartIndexZero = true;
table.Rows.Add(differentRow);
SemanticallyEquivalentCrossoverDataTable = table;
}
List<int> semanticallyEquivalentCrossoverCount = new List<int>() { 0, 0, 0 };
for (int i = 0; i < semanticallyEquivalentCrossover.Length; i++) {
semanticallyEquivalentCrossoverCount[semanticallyEquivalentCrossover[i].Value]++;
}
double total = semanticallyEquivalentCrossover.Length;
SemanticallyEquivalentCrossoverDataTable.Rows["No Crossover"].Values.Add(semanticallyEquivalentCrossoverCount[0] / total * 100.0);
SemanticallyEquivalentCrossoverDataTable.Rows["Equivalent"].Values.Add(semanticallyEquivalentCrossoverCount[1] / total * 100.0);
SemanticallyEquivalentCrossoverDataTable.Rows["Equivalent + No Crossover"].Values.Add((semanticallyEquivalentCrossoverCount[0] + semanticallyEquivalentCrossoverCount[1]) / total * 100.0);
SemanticallyEquivalentCrossoverDataTable.Rows["Different"].Values.Add(semanticallyEquivalentCrossoverCount[2] / total * 100.0);
}
public override IOperation Apply() {
UpdateCounter.Value++;
// the analyzer runs periodically, every 'updateInterval' times
if (UpdateCounter.Value == UpdateInterval.Value) {
UpdateCounter.Value = 0; // reset counter
// compute all tree lengths and store them in the lengthsTable
var solutions = SymbolicExpressionTreeParameter.ActualValue;
var treeLengthsTable = SymbolicExpressionTreeLengthsParameter.ActualValue;
// if the table was not created yet, we create it here
if (treeLengthsTable == null) {
treeLengthsTable = new DataTable("Tree Length Histogram");
SymbolicExpressionTreeLengthsParameter.ActualValue = treeLengthsTable;
}
// data table which stores tree length values
DataRow treeLengthsTableRow;
const string treeLengthsTableRowName = "Symbolic expression tree lengths";
const string treeLengthsTableRowDesc = "The distribution of symbolic expression tree lengths";
const string xAxisTitle = "Symbolic expression tree lengths";
const string yAxisTitle = "Frequency / Number of tree individuals";
var treeLengths = solutions.Select(s => (int)s.Length).ToList();
int maxLength = treeLengths.Max(t => t);
int minLength = treeLengths.Min(t => t);
if (!treeLengthsTable.Rows.ContainsKey(treeLengthsTableRowName)) {
treeLengthsTableRow = new DataRow(treeLengthsTableRowName, treeLengthsTableRowDesc, treeLengths.Select(x => (double)x));
treeLengthsTable.Rows.Add(treeLengthsTableRow);
} else {
treeLengthsTableRow = treeLengthsTable.Rows[treeLengthsTableRowName];
treeLengthsTableRow.Values.Replace(treeLengths.Select(x => (double)x));
}
double maximumAllowedTreeLength = MaximumSymbolicExpressionTreeLengthParameter.ActualValue.Value;
treeLengthsTableRow.VisualProperties.ChartType = DataRowVisualProperties.DataRowChartType.Histogram;
treeLengthsTableRow.VisualProperties.ExactBins = false;
int range = maxLength - minLength;
if (range == 0) range = 1;
// the following trick should result in an integer intervalWidth of 1,2,4,...
treeLengthsTableRow.VisualProperties.Bins = range;
if (maxLength <= 25) // [0,25]
treeLengthsTableRow.VisualProperties.ScaleFactor = 1.0;
else if (maxLength <= 100) // [26,100]
treeLengthsTableRow.VisualProperties.ScaleFactor = 1.0 / 2.0;
else if (maxLength <= 250) // [101,250]
treeLengthsTableRow.VisualProperties.ScaleFactor = 1.0 / 5.0;
else if (maxLength <= 500) // [251,500]
treeLengthsTableRow.VisualProperties.ScaleFactor = 1.0 / 10.0;
else
treeLengthsTableRow.VisualProperties.ScaleFactor = 1.0 / 20.0; // [501,inf]
treeLengthsTableRow.VisualProperties.IsVisibleInLegend = false;
// visual properties for the X-axis
treeLengthsTable.VisualProperties.XAxisMinimumAuto = false;
treeLengthsTable.VisualProperties.XAxisMaximumAuto = false;
treeLengthsTable.VisualProperties.XAxisMinimumFixedValue = 0.0;
if (maxLength > maximumAllowedTreeLength + 1)
treeLengthsTable.VisualProperties.XAxisMaximumFixedValue = maxLength + 1; // +1 so the histogram column for the maximum length won't get trimmed
else
treeLengthsTable.VisualProperties.XAxisMaximumFixedValue = maximumAllowedTreeLength + 1;
treeLengthsTable.VisualProperties.XAxisTitle = xAxisTitle;
//visual properties for the Y-axis
treeLengthsTable.VisualProperties.YAxisMinimumAuto = false;
treeLengthsTable.VisualProperties.YAxisMaximumAuto = false;
treeLengthsTable.VisualProperties.YAxisMinimumFixedValue = 0.0;
int maxFreq = (int)Math.Round(solutions.GroupBy(s => s.Length).Max(g => g.Count()) / treeLengthsTableRow.VisualProperties.ScaleFactor);
if (maxFreq % 5 != 0)
maxFreq += (5 - maxFreq % 5);
double yAxisMaximumFixedValue = maxFreq;
treeLengthsTable.VisualProperties.YAxisMaximumFixedValue = yAxisMaximumFixedValue;
treeLengthsTable.VisualProperties.YAxisTitle = yAxisTitle;
var results = ResultsParameter.ActualValue;
if (!results.ContainsKey(treeLengthsTableRowName)) {
results.Add(new Result(treeLengthsTableRowName, treeLengthsTable));
} else {
results[treeLengthsTableRowName].Value = treeLengthsTable;
}
bool storeHistory = StoreHistoryParameter.Value.Value;
const string treeLengthHistoryTableName = "Tree lengths history";
if (storeHistory) {
var treeLengthsHistory = SymbolicExpressionTreeLengthsHistoryParameter.ActualValue;
if (treeLengthsHistory == null) {
treeLengthsHistory = new DataTableHistory();
SymbolicExpressionTreeLengthsHistoryParameter.ActualValue = treeLengthsHistory;
}
treeLengthsHistory.Add((DataTable)treeLengthsTable.Clone());
if (!results.ContainsKey(treeLengthHistoryTableName)) {
results.Add(new Result(treeLengthHistoryTableName, treeLengthsHistory));
} else {
results[treeLengthHistoryTableName].Value = treeLengthsHistory;
}
}
}
return base.Apply();
}
public override IOperation Apply() {
if (SimilarityParameter.ActualValue == null || SimilarityParameter.ActualValue.Value.IsAlmost(-1.0)) {
UpdateCounter.Value++;
if (UpdateCounter.Value != UpdateInterval.Value) return base.Apply();
UpdateCounter.Value = 0;
var trees = SymbolicExpressionTreeParameter.ActualValue.ToList();
SimilarityParameter.ActualValue = new DoubleValue();
var operations = new OperationCollection { Parallel = true };
foreach (var tree in trees) {
var op = new SymbolicDataAnalysisExpressionTreeSimilarityCalculator(DistanceCalculatorParameter.Value) {
CurrentSymbolicExpressionTree = tree,
MaximumTreeDepth = MaximumSymbolicExpressionTreeDepth.Value
};
var operation = ExecutionContext.CreateChildOperation(op, ExecutionContext.Scope);
operations.Add(operation);
}
return new OperationCollection { operations, ExecutionContext.CreateOperation(this) };
}
var results = ResultsParameter.ActualValue;
// population diversity
DataTable populationDiversityTable;
if (!results.ContainsKey("PopulationDiversity")) {
populationDiversityTable = new DataTable("PopulationDiversity") { VisualProperties = { YAxisTitle = "Diversity" } };
results.Add(new Result("PopulationDiversity", populationDiversityTable));
}
populationDiversityTable = (DataTable)results["PopulationDiversity"].Value;
if (!populationDiversityTable.Rows.ContainsKey("Diversity"))
populationDiversityTable.Rows.Add(new DataRow("Diversity") { VisualProperties = { StartIndexZero = true } });
int length = SymbolicExpressionTreeParameter.ActualValue.Length;
var similarity = SimilarityParameter.ActualValue.Value / (length * (length - 1) / 2.0);
var diversity = 1 - similarity;
SimilarityParameter.ActualValue.Value = -1.0;
populationDiversityTable.Rows["Diversity"].Values.Add(diversity);
return base.Apply();
}
private void AddValues(double[] x, DataTable dataTable) {
if (!dataTable.Rows.Any()) {
for (int i = 0; i < x.Length; i++) {
var newRow = new DataRow("x" + i);
newRow.Values.Add(x[i]);
dataTable.Rows.Add(newRow);
}
} else {
for (int i = 0; i < x.Length; i++) {
dataTable.Rows.ElementAt(i).Values.Add(x[i]);
}
}
}
private void AddTypeSelectedForSimilarityTableEntry(StringValue[] typeSelectedForSimilarity) {
if (!ResultCollection.ContainsKey(TypeSelectedForSimilarityParameterName)) {
var newTable = new DataTable(TypeSelectedForSimilarityParameterName, "");
newTable.VisualProperties.YAxisTitle = "Percentage";
newTable.VisualProperties.YAxisMaximumAuto = false;
ResultCollection.Add(new Result(TypeSelectedForSimilarityParameterName, newTable));
}
var table = ((DataTable)ResultCollection[TypeSelectedForSimilarityParameterName].Value);
// all rows must have the same number of values so we can just take the first
int numberOfValues = table.Rows.Select(r => r.Values.Count).DefaultIfEmpty().First();
double count = typeSelectedForSimilarity.Count();
var groupedValues = typeSelectedForSimilarity.Select(x => x.Value).GroupBy(x => x);
foreach (var type in groupedValues) {
if (!table.Rows.ContainsKey(type.Key)) {
// initialize a new row for the symbol and pad with zeros
DataRow row = new DataRow(type.Key, "", Enumerable.Repeat(0.0, numberOfValues));
row.VisualProperties.StartIndexZero = true;
table.Rows.Add(row);
}
table.Rows[type.Key].Values.Add(type.Count() / count * 100);
}
// add a zero for each data row that was not modified in the previous loop
foreach (var row in table.Rows.Where(r => r.Values.Count != numberOfValues + 1))
row.Values.Add(0.0);
}
