Changeset 16141 for branches/2817-BinPackingSpeedup/HeuristicLab.Problems.DataAnalysis

Timestamp:

09/14/18 11:47:37 (6 years ago)

Author:

abeham

Message:

#2817: updated to trunk r16140

Location:

branches/2817-BinPackingSpeedup

Files:

• . (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis/3.4/Dataset.cs (modified) (6 diffs)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ClassificationSolutionVariableImpactsCalculator.cs (modified) (6 diffs)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/DataAnalysisProblemData.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Regression/RegressionSolutionVariableImpactsCalculator.cs (modified) (4 diffs)
• HeuristicLab.Problems.DataAnalysis/3.4/ModifiableDataset.cs (modified) (5 diffs)

branches/2817-BinPackingSpeedup/HeuristicLab.Problems.DataAnalysis/3.4/Dataset.cs

@@ -38 +38 @@
     protected Dataset(Dataset original, Cloner cloner)
       : base(original, cloner) {
+      // no need to clone the variable values because these can't be modified
       variableValues = new Dictionary<string, IList>(original.variableValues);
       variableNames = new List<string>(original.variableNames);
       rows = original.rows;
     }
+
     public override IDeepCloneable Clone(Cloner cloner) { return new Dataset(this, cloner); }
 
@@ -58 +60 @@
     /// <param name="variableValues">The values for the variables (column-oriented storage). Values are not cloned!</param>
     public Dataset(IEnumerable<string> variableNames, IEnumerable<IList> variableValues)
-      : base() {
+      : this(variableNames, variableValues, cloneValues: true) {
+    }
+
+    protected Dataset(IEnumerable<string> variableNames, IEnumerable<IList> variableValues, bool cloneValues = false) {
       Name = "-";
-      if (!variableNames.Any()) {
+
+      if (variableNames.Any()) {
+        this.variableNames = new List<string>(variableNames);
+      } else {
         this.variableNames = Enumerable.Range(0, variableValues.Count()).Select(x => "Column " + x).ToList();
-      } else if (variableNames.Count() != variableValues.Count()) {
-        throw new ArgumentException("Number of variable names doesn't match the number of columns of variableValues");
-      } else if (!variableValues.All(list => list.Count == variableValues.First().Count)) {
-        throw new ArgumentException("The number of values must be equal for every variable");
-      } else if (variableNames.Distinct().Count() != variableNames.Count()) {
-        var duplicateVariableNames =
-          variableNames.GroupBy(v => v).Where(g => g.Count() > 1).Select(g => g.Key).ToList();
-        string message = "The dataset cannot contain duplicate variables names: " + Environment.NewLine;
-        foreach (var duplicateVariableName in duplicateVariableNames)
-          message += duplicateVariableName + Environment.NewLine;
-        throw new ArgumentException(message);
-      }
+      }
+      // check if the arguments are consistent (no duplicate variables, same number of rows, correct data types, ...)
+      CheckArguments(this.variableNames, variableValues);
+
       rows = variableValues.First().Count;
-      this.variableNames = new List<string>(variableNames);
-      this.variableValues = new Dictionary<string, IList>(this.variableNames.Count);
-      for (int i = 0; i < this.variableNames.Count; i++) {
-        var values = variableValues.ElementAt(i);
-        this.variableValues.Add(this.variableNames[i], values);
+
+      if (cloneValues) {
+        this.variableValues = CloneValues(this.variableNames, variableValues);
+      } else {
+        this.variableValues = new Dictionary<string, IList>(this.variableNames.Count);
+        for (int i = 0; i < this.variableNames.Count; i++) {
+          var variableName = this.variableNames[i];
+          var values = variableValues.ElementAt(i);
+          this.variableValues.Add(variableName, values);
+        }
       }
     }
@@ -111 +116 @@
 
     public ModifiableDataset ToModifiable() {
-      var values = new List<IList>();
-      foreach (var v in variableNames) {
-        if (VariableHasType<double>(v)) {
-          values.Add(new List<double>((IList<double>)variableValues[v]));
-        } else if (VariableHasType<string>(v)) {
-          values.Add(new List<string>((IList<string>)variableValues[v]));
-        } else if (VariableHasType<DateTime>(v)) {
-          values.Add(new List<DateTime>((IList<DateTime>)variableValues[v]));
-        } else {
-          throw new ArgumentException("Unknown variable type.");
-        }
-      }
-      return new ModifiableDataset(variableNames, values);
-    }
+      return new ModifiableDataset(variableNames, variableNames.Select(v => variableValues[v]), true);
+    }
+
     /// <summary>
     /// Shuffle a dataset's rows
@@ -135 +129 @@
     }
 
-    protected Dataset(Dataset dataset) : this(dataset.variableNames, dataset.variableValues.Values) { }
+
 
     #region Backwards compatible code, remove with 3.5
@@ -231 +225 @@
       return new ReadOnlyCollection<DateTime>(values);
     }
-
-
     private IEnumerable<T> GetValues<T>(string variableName, IEnumerable<int> rows) {
       var values = GetValues<T>(variableName);
@@ -248 +240 @@
       return variableValues[variableName] is IList<T>;
     }
+    protected Type GetVariableType(string variableName) {
+      IList list;
+      variableValues.TryGetValue(variableName, out list);
+      if (list == null)
+        throw new ArgumentException("The variable " + variableName + " does not exist in the dataset.");
+      return GetElementType(list);
+    }
+    protected static Type GetElementType(IList list) {
+      var type = list.GetType();
+      return type.IsGenericType ? type.GetGenericArguments()[0] : type.GetElementType();
+    }
+    protected static bool IsAllowedType(IList list) {
+      var type = GetElementType(list);
+      return IsAllowedType(type);
+    }
+    protected static bool IsAllowedType(Type type) {
+      return type == typeof(double) || type == typeof(string) || type == typeof(DateTime);
+    }
+
+    protected static void CheckArguments(IEnumerable<string> variableNames, IEnumerable<IList> variableValues) {
+      if (variableNames.Count() != variableValues.Count()) {
+        throw new ArgumentException("Number of variable names doesn't match the number of columns of variableValues");
+      } else if (!variableValues.All(list => list.Count == variableValues.First().Count)) {
+        throw new ArgumentException("The number of values must be equal for every variable");
+      } else if (variableNames.Distinct().Count() != variableNames.Count()) {
+        var duplicateVariableNames =
+          variableNames.GroupBy(v => v).Where(g => g.Count() > 1).Select(g => g.Key).ToList();
+        string message = "The dataset cannot contain duplicate variables names: " + Environment.NewLine;
+        foreach (var duplicateVariableName in duplicateVariableNames)
+          message += duplicateVariableName + Environment.NewLine;
+        throw new ArgumentException(message);
+      }
+      // check if all the variables are supported
+      foreach (var t in variableNames.Zip(variableValues, Tuple.Create)) {
+        var variableName = t.Item1;
+        var values = t.Item2;
+
+        if (!IsAllowedType(values)) {
+          throw new ArgumentException(string.Format("Unsupported type {0} for variable {1}.", GetElementType(values), variableName));
+        }
+      }
+    }
+
+    protected static Dictionary<string, IList> CloneValues(Dictionary<string, IList> variableValues) {
+      return variableValues.ToDictionary(x => x.Key, x => CloneValues(x.Value));
+    }
+
+    protected static Dictionary<string, IList> CloneValues(IEnumerable<string> variableNames, IEnumerable<IList> variableValues) {
+      return variableNames.Zip(variableValues, Tuple.Create).ToDictionary(x => x.Item1, x => CloneValues(x.Item2));
+    }
+
+    protected static IList CloneValues(IList values) {
+      var doubleValues = values as IList<double>;
+      if (doubleValues != null) return new List<double>(doubleValues);
+
+      var stringValues = values as IList<string>;
+      if (stringValues != null) return new List<string>(stringValues);
+
+      var dateTimeValues = values as IList<DateTime>;
+      if (dateTimeValues != null) return new List<DateTime>(dateTimeValues);
+
+      throw new ArgumentException(string.Format("Unsupported variable type {0}.", GetElementType(values)));
+    }
 
     #region IStringConvertibleMatrix Members
     [Storable]
-    protected int rows;
+    private int rows;
     public int Rows {
       get { return rows; }
+      protected set { rows = value; }
     }
     int IStringConvertibleMatrix.Rows {

branches/2817-BinPackingSpeedup/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ClassificationSolutionVariableImpactsCalculator.cs

@@ -100 +100 @@
       var problemData = solution.ProblemData;
       var dataset = problemData.Dataset;
+      var model = (IClassificationModel)solution.Model.Clone(); //mkommend: clone of model is necessary, because the thresholds for IDiscriminantClassificationModels are updated
 
       IEnumerable<int> rows;
@@ -137 +138 @@
       // calculate impacts for double variables
       foreach (var inputVariable in allowedInputVariables.Where(problemData.Dataset.VariableHasType<double>)) {
-        var newEstimates = EvaluateModelWithReplacedVariable(solution.Model, inputVariable, modifiableDataset, rows, replacementMethod);
+        var newEstimates = EvaluateModelWithReplacedVariable(model, inputVariable, modifiableDataset, rows, replacementMethod);
         var newAccuracy = OnlineAccuracyCalculator.Calculate(targetValues, newEstimates, out error);
         if (error != OnlineCalculatorError.None) throw new InvalidOperationException("Error during R² calculation with replaced inputs.");
@@ -150 +151 @@
           var smallestImpact = double.PositiveInfinity;
           foreach (var repl in problemData.Dataset.GetStringValues(inputVariable, rows).Distinct()) {
-            var newEstimates = EvaluateModelWithReplacedVariable(solution.Model, inputVariable, modifiableDataset, rows,
+            var newEstimates = EvaluateModelWithReplacedVariable(model, inputVariable, modifiableDataset, rows,
               Enumerable.Repeat(repl, dataset.Rows));
             var newAccuracy = OnlineAccuracyCalculator.Calculate(targetValues, newEstimates, out error);
@@ -164 +165 @@
           // calculate impacts for factor variables
 
-          var newEstimates = EvaluateModelWithReplacedVariable(solution.Model, inputVariable, modifiableDataset, rows,
+          var newEstimates = EvaluateModelWithReplacedVariable(model, inputVariable, modifiableDataset, rows,
             factorReplacementMethod);
           var newAccuracy = OnlineAccuracyCalculator.Calculate(targetValues, newEstimates, out error);
@@ -263 +264 @@
       var originalValues = dataset.GetReadOnlyDoubleValues(variable).ToList();
       dataset.ReplaceVariable(variable, replacementValues.ToList());
+
+      var discModel = model as IDiscriminantFunctionClassificationModel;
+      if (discModel != null) {
+        var problemData = new ClassificationProblemData(dataset, dataset.VariableNames, model.TargetVariable);
+        discModel.RecalculateModelParameters(problemData, rows);
+      }
+
       //mkommend: ToList is used on purpose to avoid lazy evaluation that could result in wrong estimates due to variable replacements
       var estimates = model.GetEstimatedClassValues(dataset, rows).ToList();
@@ -273 +281 @@
       var originalValues = dataset.GetReadOnlyStringValues(variable).ToList();
       dataset.ReplaceVariable(variable, replacementValues.ToList());
+
+
+      var discModel = model as IDiscriminantFunctionClassificationModel;
+      if (discModel != null) {
+        var problemData = new ClassificationProblemData(dataset, dataset.VariableNames, model.TargetVariable);
+        discModel.RecalculateModelParameters(problemData, rows);
+      }
+
       //mkommend: ToList is used on purpose to avoid lazy evaluation that could result in wrong estimates due to variable replacements
       var estimates = model.GetEstimatedClassValues(dataset, rows).ToList();

branches/2817-BinPackingSpeedup/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/DataAnalysisProblemData.cs

@@ -163 +163 @@
 
       var variables = dataset.VariableNames.Where(variable => dataset.VariableHasType<double>(variable) || dataset.VariableHasType<string>(variable));
-      var inputVariables = new CheckedItemList<StringValue>(variables.Select(x => new StringValue(x)));
+      var inputVariables = new CheckedItemList<StringValue>(variables.Select(x => new StringValue(x).AsReadOnly()));
       foreach (StringValue x in inputVariables)
         inputVariables.SetItemCheckedState(x, allowedInputVariables.Contains(x.Value));

branches/2817-BinPackingSpeedup/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Regression/RegressionSolutionVariableImpactsCalculator.cs

@@ -52 +52 @@
       All
     }
-   
+
     private const string ReplacementParameterName = "Replacement Method";
     private const string DataPartitionParameterName = "DataPartition";
@@ -96 +96 @@
       DataPartitionEnum data = DataPartitionEnum.Training,
       ReplacementMethodEnum replacementMethod = ReplacementMethodEnum.Median,
-      FactorReplacementMethodEnum factorReplacementMethod = FactorReplacementMethodEnum.Best) {
+      FactorReplacementMethodEnum factorReplacementMethod = FactorReplacementMethodEnum.Best,
+      Func<double, string, bool> progressCallback = null) {
 
       var problemData = solution.ProblemData;
@@ -134 +135 @@
       var allowedInputVariables = dataset.VariableNames.Where(v => inputvariables.Contains(v)).ToList();
 
+      int curIdx = 0;
+      int count = allowedInputVariables.Where(problemData.Dataset.VariableHasType<double>).Count();
       // calculate impacts for double variables
       foreach (var inputVariable in allowedInputVariables.Where(problemData.Dataset.VariableHasType<double>)) {
+        //Report the current progress in percent. If the callback returns true, it means the execution shall be stopped
+        if (progressCallback != null) {
+          curIdx++;
+          if (progressCallback((double)curIdx / count, string.Format("Calculating impact for variable {0} ({1} of {2})", inputVariable, curIdx, count))) { return null; }
+        }
         var newEstimates = EvaluateModelWithReplacedVariable(solution.Model, inputVariable, modifiableDataset, rows, replacementMethod);
         var newR2 = OnlinePearsonsRCalculator.Calculate(targetValues, newEstimates, out error);
@@ -180 +188 @@
     }
 
+
     private static IEnumerable<double> EvaluateModelWithReplacedVariable(IRegressionModel model, string variable, ModifiableDataset dataset, IEnumerable<int> rows, ReplacementMethodEnum replacement = ReplacementMethodEnum.Median) {
       var originalValues = dataset.GetReadOnlyDoubleValues(variable).ToList();

branches/2817-BinPackingSpeedup/HeuristicLab.Problems.DataAnalysis/3.4/ModifiableDataset.cs

@@ -39 +39 @@
 
     private ModifiableDataset(ModifiableDataset original, Cloner cloner) : base(original, cloner) {
-      var variables = variableValues.Keys.ToList();
-      foreach (var v in variables) {
-        var type = GetVariableType(v);
-        if (type == typeof(DateTime)) {
-          variableValues[v] = GetDateTimeValues(v).ToList();
-        } else if (type == typeof(double)) {
-          variableValues[v] = GetDoubleValues(v).ToList();
-        } else if (type == typeof(string)) {
-          variableValues[v] = GetStringValues(v).ToList();
-        } else {
-          throw new ArgumentException("Unsupported type " + type + " for variable " + v);
+      variableNames = new List<string>(original.variableNames);
+      variableValues = CloneValues(original.variableValues);
+    }
+
+    public override IDeepCloneable Clone(Cloner cloner) { return new ModifiableDataset(this, cloner); }
+
+    public ModifiableDataset() { }
+
+    public ModifiableDataset(IEnumerable<string> variableNames, IEnumerable<IList> variableValues, bool cloneValues = false) :
+      base(variableNames, variableValues, cloneValues) { }
+
+    public Dataset ToDataset() {
+      return new Dataset(variableNames, variableNames.Select(v => variableValues[v]));
+    }
+
+
+    public IEnumerable<object> GetRow(int row) {
+      return variableValues.Select(x => x.Value[row]);
+    }
+
+    public void AddRow(IEnumerable<object> values) {
+      var list = values.ToList();
+      if (list.Count != variableNames.Count)
+        throw new ArgumentException("The number of values must be equal to the number of variable names.");
+      // check if all the values are of the correct type
+      for (int i = 0; i < list.Count; ++i) {
+        if (list[i].GetType() != GetVariableType(variableNames[i])) {
+          throw new ArgumentException("The type of the provided value does not match the variable type.");
         }
       }
-    }
-    public override IDeepCloneable Clone(Cloner cloner) { return new ModifiableDataset(this, cloner); }
-    public ModifiableDataset() : base() { }
-
-    public ModifiableDataset(IEnumerable<string> variableNames, IEnumerable<IList> variableValues) : base(variableNames, variableValues) { }
+      // add values
+      for (int i = 0; i < list.Count; ++i) {
+        variableValues[variableNames[i]].Add(list[i]);
+      }
+      Rows++;
+      OnRowsChanged();
+      OnReset();
+    }
 
     public void ReplaceRow(int row, IEnumerable<object> values) {
@@ -72 +92 @@
         variableValues[variableNames[i]][row] = list[i];
       }
+      OnReset();
+    }
+
+    // slow, avoid using this
+    public void RemoveRow(int row) {
+      foreach (var list in variableValues.Values)
+        list.RemoveAt(row);
+      Rows--;
+      OnRowsChanged();
+      OnReset();
+    }
+
+    // adds a new variable to the dataset
+    public void AddVariable(string variableName, IList values) {
+      InsertVariable(variableName, Columns, values);
+    }
+
+    public void InsertVariable(string variableName, int position, IList values) {
+      if (variableValues.ContainsKey(variableName))
+        throw new ArgumentException(string.Format("Variable {0} is already present in the dataset.", variableName));
+
+      if (position < 0 || position > Columns)
+        throw new ArgumentException(string.Format("Incorrect position {0} specified. The position must be between 0 and {1}.", position, Columns));
+
+      if (values == null)
+        throw new ArgumentNullException("values", "Values must not be null. At least an empty list of values has to be provided.");
+
+      if (values.Count != Rows)
+        throw new ArgumentException(string.Format("{0} values are provided, but {1} rows are present in the dataset.", values.Count, Rows));
+
+      if (!IsAllowedType(values))
+        throw new ArgumentException(string.Format("Unsupported type {0} for variable {1}.", GetElementType(values), variableName));
+
+      variableNames.Insert(position, variableName);
+      variableValues[variableName] = values;
+
+      OnColumnsChanged();
+      OnColumnNamesChanged();
       OnReset();
     }
@@ -85 +143 @@
     }
 
-    public void AddRow(IEnumerable<object> values) {
-      var list = values.ToList();
-      if (list.Count != variableNames.Count)
-        throw new ArgumentException("The number of values must be equal to the number of variable names.");
-      // check if all the values are of the correct type
-      for (int i = 0; i < list.Count; ++i) {
-        if (list[i].GetType() != GetVariableType(variableNames[i])) {
-          throw new ArgumentException("The type of the provided value does not match the variable type.");
-        }
-      }
-      // add values
-      for (int i = 0; i < list.Count; ++i) {
-        variableValues[variableNames[i]].Add(list[i]);
-      }
-      rows++;
-      OnRowsChanged();
-      OnReset();
-    }
-
-    // adds a new variable to the dataset
-    public void AddVariable<T>(string variableName, IEnumerable<T> values) {
-      if (variableValues.ContainsKey(variableName))
-        throw new ArgumentException("Variable " + variableName + " is already present in the dataset.");
-      int count = values.Count();
-      if (count != rows)
-        throw new ArgumentException("The number of values must exactly match the number of rows in the dataset.");
-      variableValues[variableName] = new List<T>(values);
-      variableNames.Add(variableName);
-      OnColumnsChanged();
-      OnColumnNamesChanged();
-      OnReset();
-    }
 
     public void RemoveVariable(string variableName) {
       if (!variableValues.ContainsKey(variableName))
-        throw new ArgumentException("The variable " + variableName + " does not exist in the dataset.");
+        throw new ArgumentException(string.Format("The variable {0} does not exist in the dataset.", variableName));
       variableValues.Remove(variableName);
       variableNames.Remove(variableName);
@@ -128 +154 @@
     }
 
-    // slow, avoid to use this
-    public void RemoveRow(int row) {
-      foreach (var list in variableValues.Values)
-        list.RemoveAt(row);
-      rows--;
+    public void ClearValues() {
+      foreach (var list in variableValues.Values) {
+        list.Clear();
+      }
+      Rows = 0;
       OnRowsChanged();
       OnReset();
     }
+
 
     public void SetVariableValue(object value, string variableName, int row) {
@@ -151 +178 @@
     }
 
-    private Type GetVariableType(string variableName) {
-      IList list;
-      variableValues.TryGetValue(variableName, out list);
-      if (list == null)
-        throw new ArgumentException("The variable " + variableName + " does not exist in the dataset.");
-      return list.GetType().GetGenericArguments()[0];
-    }
-
     bool IStringConvertibleMatrix.SetValue(string value, int rowIndex, int columnIndex) {
       var variableName = variableNames[columnIndex];