Changeset 9121 for branches/CMAES/HeuristicLab.Encodings.RealVectorEncoding

Timestamp:

01/07/13 20:21:07 (12 years ago)

Author:

abeham

Message:

#1961: Added wiring code, removed obsolete parameters, simplified mainloop slightly, changed sigmabounds to a matrix

Location:

branches/CMAES/HeuristicLab.Encodings.RealVectorEncoding/3.3/CMAESOperators

Files:

• CMAInitializer.cs (modified) (4 diffs)
• CMAMutator.cs (modified) (2 diffs)
• CMAParameters.cs (modified) (4 diffs)
• CMAUpdater.cs (modified) (1 diff)

branches/CMAES/HeuristicLab.Encodings.RealVectorEncoding/3.3/CMAESOperators/CMAInitializer.cs

@@ -48 +48 @@
     }
 
-    public IValueLookupParameter<DoubleArray> MaxSigmaParameter {
-      get { return (IValueLookupParameter<DoubleArray>)Parameters["MaxSigma"]; }
-    }
-
-    public IValueLookupParameter<DoubleArray> MinSigmaParameter {
-      get { return (IValueLookupParameter<DoubleArray>)Parameters["MinSigma"]; }
+    public IValueLookupParameter<DoubleMatrix> SigmaBoundsParameter {
+      get { return (IValueLookupParameter<DoubleMatrix>)Parameters["SigmaBounds"]; }
     }
 
@@ -88 +84 @@
       Parameters.Add(new ValueLookupParameter<IntValue>("Dimension", "The problem dimension (N)."));
       Parameters.Add(new ValueLookupParameter<DoubleArray>("InitialSigma", "The initial value for Sigma (need to be > 0), can be single dimensioned or an array that should be equal to the size of the vector.", new DoubleArray(new[] { 0.5 })));
-      Parameters.Add(new ValueLookupParameter<DoubleArray>("MaxSigma", "The maximum sigma value can be omitted, given as one value for all dimensions or a value for each dimension."));
-      Parameters.Add(new ValueLookupParameter<DoubleArray>("MinSigma", "The minimum sigma value can be omitted, given as one value for all dimensions or a value for each dimension."));
+      Parameters.Add(new ValueLookupParameter<DoubleMatrix>("SigmaBounds", "The bounds for sigma value can be omitted, given as one value for all dimensions or a value for each dimension. First column specifies minimum, second column maximum value."));
       Parameters.Add(new LookupParameter<IntValue>("Iterations", "The current iteration that is being processed."));
       Parameters.Add(new ValueLookupParameter<IntValue>("MaximumIterations", "The maximum number of iterations to be processed."));
-      Parameters.Add(new ValueLookupParameter<IntValue>("InitialIterations", "The number of iterations that should be performed using the diagonal covariance matrix only."));
+      Parameters.Add(new ValueLookupParameter<IntValue>("InitialIterations", "The number of iterations that should be performed using the diagonal covariance matrix only.", new IntValue(0)));
       Parameters.Add(new LookupParameter<IntValue>("PopulationSize", "The population size (lambda)."));
       Parameters.Add(new LookupParameter<IntValue>("Mu", "The number of offspring considered for updating of the strategy parameters."));
@@ -132 +127 @@
 
       // ensure maximal and minimal standard deviations
-      var minSigma = MinSigmaParameter.ActualValue;
-      if (minSigma != null && minSigma.Length > 0) {
+      var sigmaBounds = SigmaBoundsParameter.ActualValue;
+      if (sigmaBounds != null && sigmaBounds.Rows > 0) {
         for (int i = 0; i < N; i++) {
-          var d = minSigma[(int)Math.Min(i, minSigma.Length - 1)];
+          var d = sigmaBounds[(int)Math.Min(i, sigmaBounds.Rows - 1), 0];
           if (d > sigma * minSqrtdiagC) sigma = d / minSqrtdiagC;
         }
-      }
-      var maxSigma = MaxSigmaParameter.ActualValue;
-      if (maxSigma != null && maxSigma.Length > 0) {
         for (int i = 0; i < N; i++) {
-          var d = maxSigma[(int)Math.Min(i, maxSigma.Length - 1)];
+          var d = sigmaBounds[(int)Math.Min(i, sigmaBounds.Rows - 1), 1];
           if (d > sigma * maxSqrtdiagC) sigma = d / maxSqrtdiagC;
         }
@@ -181 +173 @@
       sp.BDz = new DoubleArray(BDz);
       sp.Sigma = new DoubleValue(sigma);
-      if (maxSigma != null) sp.MaxSigma = (DoubleArray)maxSigma.Clone();
-      if (minSigma != null) sp.MinSigma = (DoubleArray)minSigma.Clone();
+      if (sigmaBounds != null) sp.SigmaBounds = (DoubleMatrix)sigmaBounds.Clone();
       sp.InitialIterations = new IntValue(initialIterations.Value);
 

branches/CMAES/HeuristicLab.Encodings.RealVectorEncoding/3.3/CMAESOperators/CMAMutator.cs

@@ -105 +105 @@
       } else {
         bool inRange;
+        var B = sp.B;
         tries = 0;
         do {
@@ -116 +117 @@
             var sum = 0.0;
             for (int j = 0; j < arx.Length; j++)
-              sum += sp.B[i, j] * artmp[j];
+              sum += B[i, j] * artmp[j];
             arx[i] = copy[i] + sp.Sigma.Value * sum; // m + sig * Normal(0,C) 
             if (bounds[i % bounds.Rows, 0] > arx[i] || arx[i] > bounds[i % bounds.Rows, 1])

branches/CMAES/HeuristicLab.Encodings.RealVectorEncoding/3.3/CMAESOperators/CMAParameters.cs

@@ -56 +56 @@
 
     [Storable]
-    private DoubleArray minSigma;
-    public DoubleArray MinSigma {
-      get { return minSigma; }
-      set {
-        if (minSigma == value) return;
-        minSigma = value;
-        OnPropertyChanged("MinSigma");
-      }
-    }
-
-    [Storable]
-    private DoubleArray maxSigma;
-    public DoubleArray MaxSigma {
-      get { return maxSigma; }
-      set {
-        if (maxSigma == value) return;
-        maxSigma = value;
-        OnPropertyChanged("MaxSigma");
+    private DoubleMatrix sigmaBounds;
+    public DoubleMatrix SigmaBounds {
+      get { return sigmaBounds; }
+      set {
+        if (sigmaBounds == value) return;
+        sigmaBounds = value;
+        OnPropertyChanged("SigmaBounds");
       }
     }
@@ -129 +118 @@
         cs = value;
         OnPropertyChanged("CS");
-      }
-    }
-
-    [Storable]
-    private DoubleValue c1;
-    public DoubleValue C1 {
-      get { return c1; }
-      set {
-        if (c1 == value) return;
-        c1 = value;
-        OnPropertyChanged("C1");
-      }
-    }
-
-    [Storable]
-    private DoubleValue cmu;
-    public DoubleValue CMU {
-      get { return cmu; }
-      set {
-        if (cmu == value) return;
-        cmu = value;
-        OnPropertyChanged("CMU");
       }
     }
@@ -292 +259 @@
       this.axisRatio = cloner.Clone(original.axisRatio);
       this.b = cloner.Clone(original.b);
+      this.bDz = cloner.Clone(original.bDz);
       this.c = cloner.Clone(original.c);
-      this.c1 = cloner.Clone(original.c1);
-      this.cc = cloner.Clone(original.cc);
       this.cCov = cloner.Clone(original.cCov);
       this.cCovSep = cloner.Clone(original.cCovSep);
+      this.cc = cloner.Clone(original.cc);
       this.chiN = cloner.Clone(original.chiN);
-      this.cmu = cloner.Clone(original.cmu);
       this.cs = cloner.Clone(original.cs);
       this.d = cloner.Clone(original.d);
+      this.damps = cloner.Clone(original.damps);
       this.initialIterations = cloner.Clone(original.initialIterations);
-      this.bDz = cloner.Clone(original.bDz);
-      this.maxSigma = cloner.Clone(original.maxSigma);
-      this.minSigma = cloner.Clone(original.minSigma);
       this.mu = cloner.Clone(original.mu);
       this.muCov = cloner.Clone(original.muCov);
@@ -311 +275 @@
       this.ps = cloner.Clone(original.ps);
       this.sigma = cloner.Clone(original.sigma);
+      this.sigmaBounds = cloner.Clone(original.sigmaBounds);
       this.weights = cloner.Clone(original.weights);
     }

branches/CMAES/HeuristicLab.Encodings.RealVectorEncoding/3.3/CMAESOperators/CMAUpdater.cs

@@ -157 +157 @@
 
       // ensure maximal and minimal standard deviations
-      if (sp.MinSigma != null && sp.MinSigma.Length > 0) {
-        for (int i = 0; i < N; i++) {
-          var d = sp.MinSigma[Math.Min(i, sp.MinSigma.Length - 1)];
+      if (sp.SigmaBounds != null && sp.SigmaBounds.Rows > 0) {
+        for (int i = 0; i < N; i++) {
+          var d = sp.SigmaBounds[Math.Min(i, sp.SigmaBounds.Rows - 1), 0];
           if (d > sp.Sigma.Value * minSqrtdiagC) sp.Sigma.Value = d / minSqrtdiagC;
         }
-      }
-      if (sp.MaxSigma != null && sp.MaxSigma.Length > 0) {
-        for (int i = 0; i < N; i++) {
-          var d = sp.MaxSigma[Math.Min(i, sp.MaxSigma.Length - 1)];
+        for (int i = 0; i < N; i++) {
+          var d = sp.SigmaBounds[Math.Min(i, sp.SigmaBounds.Rows - 1), 1];
           if (d > sp.Sigma.Value * maxSqrtdiagC) sp.Sigma.Value = d / maxSqrtdiagC;
         }