Changeset 15842 for branches/2886_SymRegGrammarEnumeration/ExpressionClustering/Program.cs

Timestamp:

03/15/18 10:41:20 (6 years ago)

Author:

gkronber

Message:

#2886: added clustering of functions and output of clusters, fixed bug in evaluation

File:

• branches/2886_SymRegGrammarEnumeration/ExpressionClustering/Program.cs (modified) (10 diffs)

branches/2886_SymRegGrammarEnumeration/ExpressionClustering/Program.cs

@@ -1 +1 @@
 using System;
 using System.Collections.Generic;
+using System.Drawing;
 using System.IO;
 using System.Linq;
+using HeuristicLab.Analysis;
+using HeuristicLab.Analysis.Views;
+using System.Windows.Forms;
 
 namespace ExpressionClustering {
   class Program {
     private static readonly string folder = Environment.GetFolderPath(Environment.SpecialFolder.DesktopDirectory);
+    private static readonly string clusterFolder = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.DesktopDirectory), "clusters");
     private static readonly string distinctSentencesFileName = Path.Combine(folder, @"distinctSentences.csv");
     private static readonly string allSentencesFileName = Path.Combine(folder, "allSentences.csv");
     private static readonly string outputFileName = Path.Combine(folder, "evaluations.csv.gz");
-    private static readonly string columnDelimiter = ";";
     private static int N = 100;
     private static double[] evalBuf = new double[N];
@@ -31 +35 @@
     static void Main(string[] args) {
 
-      TestFLANN();
-
       var hash2Sentences = new Dictionary<string, List<string>>();
+      // for debugging only
+      var postfix2infix = new Dictionary<string, string>();
 
 
@@ -52 +56 @@
           }
           ls.Add(toks[1]);
+          postfix2infix.Add(toks[1], toks[0]);
           nSentences++;
         }
@@ -59 +64 @@
       }
 
-      using (var writer = new StreamWriter(new System.IO.Compression.GZipStream(new FileStream(outputFileName, FileMode.OpenOrCreate), System.IO.Compression.CompressionMode.Compress)))
-        foreach (var kvp in hash2Sentences) {
-          var ls = kvp.Value;
-          var sentence = FindShortest(ls);
-          Evaluate(sentence, xs, evalBuf);
-          Add(writer, sentence, evalBuf);
-        }
-    }
-
-    private static void TestFLANN() {
-      var rand = new Random(1234);
-      var dim = 100;
-      var N = 10000;
-      var dataset = new List<double[]>();
-      for(int i=0;i<N;i++) {
-        var x = new double[dim];
-        for (int j = 0; j < dim; j++) x[j] = rand.NextDouble();
-        dataset.Add(x);
-      }
-      List<int> nnResults;
-      List<double> nnDists;
-      Flann.FindNearestNeighbours(dataset, dataset, out nnResults, out nnDists);
-
-    }
+      List<double[]> functions = new List<double[]>();
+      List<string> sentences = new List<string>();
+      List<double> qualities = new List<double>();
+
+      foreach (var kvp in hash2Sentences) {
+        var ls = kvp.Value;
+        var sentence = FindShortest(ls);
+        Evaluate(sentence, xs, evalBuf);
+        if (evalBuf.Any(ei => float.IsInfinity((float)ei) || float.IsNaN((float)ei))) {
+          Console.WriteLine("skipping {0} {1}", evalBuf.Average(), sentence);
+        } else {
+          try {
+            Scale(evalBuf, ys);
+            functions.Add((double[])evalBuf.Clone());
+            sentences.Add(sentence);
+            HeuristicLab.Problems.DataAnalysis.OnlineCalculatorError error;
+            qualities.Add(HeuristicLab.Problems.DataAnalysis.OnlinePearsonsRSquaredCalculator.Calculate(evalBuf, ys, out error));
+          } catch (ArgumentException e) {
+            // scaling failed
+          }
+        }
+      }
+
+      List<int> clusters;
+      List<double> distances;
+      Flann.FindClusters(functions, out clusters, out distances, 100);
+      
+      // output all clusters and functions
+      using (var writer = new StreamWriter(new System.IO.Compression.GZipStream(new FileStream(outputFileName, FileMode.OpenOrCreate), System.IO.Compression.CompressionMode.Compress))) {
+        for (int i = 0; i < functions.Count; i++) {
+          writer.WriteLine("{0};{1};{2};{3};{4};{5}", clusters[i], distances[i], qualities[i], sentences[i], postfix2infix[sentences[i]], string.Join(";", functions[i].Select(fi => fi.ToString())));
+        }
+      }
+
+      var funClusters = functions.Zip(clusters, (f, c) => Tuple.Create(f, c)).GroupBy(t => t.Item2);
+      var dtView = new DataTableView();
+      dtView.Size = new Size(800, 600);
+
+      foreach (var funCluster in funClusters) {
+        // draw the functions for each cluster into a separate png
+        var dtName = string.Format("R² {0}", Enumerable.Range(0, qualities.Count).Where(idx => clusters[idx] == funCluster.Key).Select(idx => qualities[idx]).Average());
+        var dt = new DataTable(dtName, dtName);
+        var rows = new List<DataRow>();
+        int i = 0;
+        foreach (var fun in funCluster.Select(t => t.Item1)) {
+          var name = i.ToString();
+          var dr = new DataRow(name, name, fun);
+          rows.Add(dr);
+          i++;
+        }
+        dt.Rows.AddRange(rows);
+        dtView.Content = dt;
+        using (var bm = new Bitmap(800, 600)) {
+          dtView.DrawToBitmap(bm, new Rectangle(0, 0, 800, 600));
+          bm.Save(Path.Combine(clusterFolder, string.Format("cluster_{0,3}.png", funCluster.Key)));
+        }
+      }
+    }
+
+
 
     private static string FindShortest(List<string> ls) {
@@ -91 +132 @@
       return minElem;
     }
+
 
 
@@ -102 +144 @@
       HeuristicLab.Problems.DataAnalysis.OnlineLinearScalingParameterCalculator.Calculate(evalBuf, ys, out alpha, out beta, out error);
       if (error != HeuristicLab.Problems.DataAnalysis.OnlineCalculatorError.None) {
-        alpha = 0.0;
-        beta = 1.0;
-        Console.WriteLine("WARNING: error in scaling");
+        throw new ArgumentException();
       }
 
@@ -110 +150 @@
         evalBuf[i] = beta * evalBuf[i] + alpha;
       }
-
-
-      // 
-      //   var meanE = 0.0;
-      //   var meanE2 = 0.0;
-      //   var meanY = 0.0;
-      //   var meanY2 = 0.0;
-      //   for(int i=0;i<evalBuf.Length;i++) {
-      //     var deltaE = evalBuf[i] - meanE;
-      //     meanE +=  deltaE / (i+1);
-      //     var deltaE2 = evalBuf[i] - meanE;
-      //     meanE2 += deltaE * deltaE2;
-      // 
-      // 
-      //     var deltaY = ys[i] - meanY;
-      //     meanY += deltaY / (i + 1);
-      //     var deltaY2 = ys[i] - meanY;
-      //     meanY2 += deltaY * deltaY2;
-      //     TODO COVARIANCE
-      //     Linear Scaling: b = cov(y,e) / var(e); a = meanY - b*meanE; 
-      //   }
-      // 
-      //   var varE = meanE2 / evalBuf.Length;
-      //   var varY = meanY2 / evalBuf.Length;
-      //   Console.WriteLine("{0} {1} {2} {3}", meanE, evalBuf.Average(), meanY, ys.Average());
-      //   Console.WriteLine("{0} {1}", varE, varY);
-      // 
-      //   var factor = varY / varE;
-      //   for(int i=0;i<evalBuf.Length;i++) {
-      //     evalBuf[i] = (evalBuf[i] - meanE) * factor + meanY;
-      //   }
     }
 
@@ -158 +167 @@
               var b = stack[topOfStack - 1];
               for (int i = 0; i < N; i++) {
-                a[i] += b[i];
-              }
+                b[i] += a[i];
+              }
+              topOfStack--;
               break;
             }
@@ -167 +177 @@
               var b = stack[topOfStack - 1];
               for (int i = 0; i < N; i++) {
-                a[i] *= b[i];
-              }
+                b[i] *= a[i];
+              }
+              topOfStack--;
               break;
             }
@@ -231 +242 @@
     }
     #endregion
-
-
-
-    // add the line with it's evaluation result to a data structure for clustering
-    private static void Add(StreamWriter writer, string line, double[] evalBuf) {
-      var avg = evalBuf.Average();
-      if (double.IsNaN(avg) || double.IsInfinity(avg)) {
-        Console.WriteLine("skipping {0} {1}", evalBuf.Average(), line);
-      } else {
-        Scale(evalBuf, ys);
-
-        writer.WriteLine(string.Join("\t", evalBuf.Select(ei => ei.ToString(System.Globalization.CultureInfo.InvariantCulture))));
-      }
-    }
   }
 }