[9430] | 1 | PROBLEM SymbReg
|
---|
| 2 |
|
---|
| 3 | CODE <<
|
---|
| 4 | double[,] inputValues;
|
---|
| 5 | double[] targetValues;
|
---|
| 6 | string[] variableNames;
|
---|
| 7 | Dictionary<string,int> nameToCol;
|
---|
| 8 |
|
---|
| 9 | double GetValue(double[,] data, string varName, int row) {
|
---|
| 10 | if(nameToCol == null) {
|
---|
| 11 | /* init mapping */
|
---|
| 12 | nameToCol = new Dictionary<string, int>();
|
---|
| 13 | for(int i=0; i<variableNames.Length; i++) {
|
---|
| 14 | nameToCol[variableNames[i]] = i;
|
---|
| 15 | }
|
---|
| 16 | }
|
---|
| 17 | return data[row, nameToCol[varName]];
|
---|
| 18 | }
|
---|
| 19 |
|
---|
| 20 | double RSquared(IEnumerable<double> xs, IEnumerable<double> ys) {
|
---|
| 21 | HeuristicLab.Problems.DataAnalysis.OnlineCalculatorError error;
|
---|
| 22 | var r2 = HeuristicLab.Problems.DataAnalysis.OnlinePearsonsRSquaredCalculator.Calculate(xs, ys, out error);
|
---|
| 23 | if(error == HeuristicLab.Problems.DataAnalysis.OnlineCalculatorError.None) return r2;
|
---|
| 24 | else return 0.0;
|
---|
| 25 | }
|
---|
| 26 |
|
---|
| 27 |
|
---|
| 28 | void LoadData(string fileName, out double[,] inputValues, out string[] variableNames, out double[] target) {
|
---|
| 29 | var prov = new HeuristicLab.Problems.Instances.DataAnalysis.RegressionRealWorldInstanceProvider();
|
---|
| 30 | var dd = prov.GetDataDescriptors().OfType<HeuristicLab.Problems.Instances.DataAnalysis.Housing>().Single();
|
---|
| 31 | var problemData = prov.LoadData(dd);
|
---|
| 32 |
|
---|
| 33 | inputValues = new double[problemData.TrainingIndices.Count(), problemData.AllowedInputVariables.Count()];
|
---|
| 34 | foreach(var r in problemData.TrainingIndices) {
|
---|
| 35 | int i=0;
|
---|
| 36 | foreach(var v in problemData.AllowedInputVariables) {
|
---|
| 37 | inputValues[r, i++] = problemData.Dataset.GetDoubleValue(v, r);
|
---|
| 38 | }
|
---|
| 39 | }
|
---|
| 40 | variableNames = problemData.AllowedInputVariables.ToArray();
|
---|
| 41 | target = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TrainingIndices).ToArray();
|
---|
| 42 | }
|
---|
| 43 | >>
|
---|
| 44 |
|
---|
| 45 | INIT <<
|
---|
| 46 | LoadData("filename.csv", out inputValues, out variableNames, out targetValues);
|
---|
| 47 | >>
|
---|
| 48 |
|
---|
| 49 | NONTERMINALS
|
---|
| 50 | Model<<int row, out double val>>.
|
---|
| 51 | RPB<<int row, out double val>>.
|
---|
| 52 | Addition<<int row, out double val>>.
|
---|
| 53 | Subtraction<<int row, out double val>>.
|
---|
| 54 | Multiplication<<int row, out double val>>.
|
---|
| 55 | Division<<int row, out double val>>.
|
---|
| 56 |
|
---|
| 57 | TERMINALS
|
---|
| 58 | Const<<out double val>>
|
---|
| 59 | CONSTRAINTS
|
---|
| 60 | val IN RANGE <<-100>> .. <<100>>
|
---|
| 61 | .
|
---|
| 62 | Var<<out string varName, out double weight>>
|
---|
| 63 | CONSTRAINTS
|
---|
| 64 | varName IN SET <<variableNames>>
|
---|
| 65 | weight IN RANGE <<-100>> .. <<100>>
|
---|
| 66 | .
|
---|
| 67 |
|
---|
| 68 | RULES
|
---|
| 69 | Model<<int row, out double val>> =
|
---|
| 70 | RPB<<row, out val>> .
|
---|
| 71 |
|
---|
| 72 | RPB<<int row, out double val>> = LOCAL << string varName; double w; >>
|
---|
| 73 | Addition<<row, out val>>
|
---|
| 74 | | Subtraction<<row, out val>>
|
---|
| 75 | | Division<<row, out val>>
|
---|
| 76 | | Multiplication<<row, out val>>
|
---|
| 77 | | Var<<out varName, out w>> SEM << val = w * GetValue(inputValues, varName, row); >>
|
---|
| 78 | | Const<<out val>>
|
---|
| 79 | .
|
---|
| 80 |
|
---|
| 81 | Addition<<int row, out double val>> = LOCAL << double x1, x2; >>
|
---|
| 82 | RPB<<row, out x1>> RPB<<row, out x2>> SEM << val = x1 + x2; >>
|
---|
| 83 | .
|
---|
| 84 | Subtraction<<int row, out double val>> = LOCAL << double x1, x2; >>
|
---|
| 85 | RPB<<row, out x1>> RPB<<row, out x2>> SEM << val = x1 - x2; >>
|
---|
| 86 | .
|
---|
| 87 | Division<<int row, out double val>> = LOCAL << double x1, x2; >>
|
---|
| 88 | RPB<<row, out x1>> RPB<<row, out x2>> SEM << val = x1 / x2; >>
|
---|
| 89 | .
|
---|
| 90 | Multiplication<<int row, out double val>> = LOCAL << double x1, x2; >>
|
---|
| 91 | RPB<<row, out x1>> RPB<<row, out x2>> SEM << val = x1 * x2; >>
|
---|
| 92 | .
|
---|
| 93 |
|
---|
| 94 | MAXIMIZE /* could also use the keyword MINIMIZE here */
|
---|
| 95 | <<
|
---|
| 96 | var rows = System.Linq.Enumerable.Range(0, inputValues.GetLength(0));
|
---|
| 97 | var predicted = rows.Select(r => {
|
---|
| 98 | double result;
|
---|
| 99 | Model(r, out result); /* we can call the root symbol directly */
|
---|
| 100 | return result;
|
---|
| 101 | });
|
---|
| 102 | return RSquared(predicted, targetValues);
|
---|
| 103 | >>
|
---|
| 104 | END SymbReg.
|
---|