Changeset 16689
 Timestamp:
 03/16/19 12:49:37 (11 months ago)
 File:

 1 edited
Legend:
 Unmodified
 Added
 Removed

branches/2974_Constants_Optimization/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/ConstantsOptimization/LMConstantsOptimizer.cs
r16514 r16689 64 64 65 65 //Variables of the symbolic expression tree correspond to parameters in the term 66 //Hence if no parameters are present no variables occur in the tree and the R² = 066 //Hence if no parameters are present we can't do anything and R² stays the same. 67 67 if (term.Parameters.Count == 0) return 0.0; 68 68 … … 103 103 int numberOfConstants = optimizedConstants.Length; 104 104 105 alglib. lsfitstate state;106 alglib. lsfitreport rep;105 alglib.minlmstate state; 106 alglib.minlmreport rep; 107 107 alglib.ndimensional_rep xrep = (p, f, obj) => LM_IterationCallback(p, f, obj); 108 int retVal;109 108 110 109 try { 111 alglib. lsfitcreatefg(x, y, optimizedConstants, numberOfRows, numberOfColumns, numberOfConstants, cheapfg: false, state: out state);112 alglib. lsfitsetcond(state, 0.0, 0.0, maxIterations);113 alglib. lsfitsetxrep(state, LM_IterationCallback != null);114 alglib. lsfitfit(state, Evaluate, EvaluateGradient, xrep, term);115 alglib. lsfitresults(state, out retVal, out optimizedConstants, out rep);110 alglib.minlmcreatevj(numberOfRows, optimizedConstants, state: out state); 111 alglib.minlmsetcond(state, 0.0, 0.0, 0.0, maxIterations); 112 alglib.minlmsetxrep(state, LM_IterationCallback != null); 113 alglib.minlmoptimize(state, Evaluate, EvaluateGradient, xrep, new object[] { term, x, y }); 114 alglib.minlmresults(state, out optimizedConstants, out rep); 116 115 } catch (ArithmeticException) { 117 116 constants = new double[0]; … … 122 121 } 123 122 123 // error 124 if (rep.terminationtype < 0) { 125 constants = initialConstants; return 0; 126 } 124 127 constants = optimizedConstants; 125 return rep.r2; 128 129 // calculate prediction with optimized constants to calculate R² 130 double[] pred = new double[numberOfRows]; 131 double[] zeros = new double[numberOfRows]; 132 Evaluate(constants, pred, new object[] { term, x, zeros }); 133 var r = OnlinePearsonsRCalculator.Calculate(pred, y, out OnlineCalculatorError error); 134 if (error != OnlineCalculatorError.None) r = 0; 135 return r * r; 126 136 } 127 137 128 129 private static void Evaluate(double[] c, double[] x, ref double fx, object o) { 130 AutoDiff.IParametricCompiledTerm term = (AutoDiff.IParametricCompiledTerm)o; 131 fx = term.Evaluate(c, x); 138 private static void Evaluate(double[] c, double[] fi, object o) { 139 var objs = (object[])o; 140 AutoDiff.IParametricCompiledTerm term = (AutoDiff.IParametricCompiledTerm)objs[0]; 141 var x = (double[,])objs[1]; 142 var y = (double[])objs[2]; 143 double[] xi = new double[x.GetLength(1)]; 144 for (int i = 0; i < fi.Length; i++) { 145 // copy data row 146 for (int j = 0; j < xi.Length; j++) xi[j] = x[i, j]; 147 fi[i] = term.Evaluate(c, xi)  y[i]; 148 } 132 149 } 133 150 134 private static void EvaluateGradient(double[] c, double[] x, ref double fx, double[] grad, object o) { 135 AutoDiff.IParametricCompiledTerm term = (AutoDiff.IParametricCompiledTerm)o; 136 Tuple<double[], double> result = term.Differentiate(c, x); 137 fx = result.Item2; 138 Array.Copy(result.Item1, grad, grad.Length); 151 private static void EvaluateGradient(double[] c, double[] fi, double[,] jac, object o) { 152 var objs = (object[])o; 153 AutoDiff.IParametricCompiledTerm term = (AutoDiff.IParametricCompiledTerm)objs[0]; 154 var x = (double[,])objs[1]; 155 var y = (double[])objs[2]; 156 double[] xi = new double[x.GetLength(1)]; 157 for (int i = 0; i < fi.Length; i++) { 158 // copy data row 159 for (int j = 0; j < xi.Length; j++) xi[j] = x[i, j]; 160 Tuple<double[], double> result = term.Differentiate(c, xi); 161 fi[i] = result.Item2  y[i]; 162 var g = result.Item1; 163 // copy gradient to Jacobian 164 for (int j = 0; j < c.Length; j++) { 165 jac[i, j] = g[j]; 166 } 167 } 139 168 } 140 169 }
Note: See TracChangeset
for help on using the changeset viewer.