YACS_ROOT_DIR variable suppressed from PMML tests

[modules/yacs.git] / src / pmml / Test / samples / unittest_ref_ann_model.cpp

diff --git a/src/pmml/Test/samples/unittest_ref_ann_model.cpp b/src/pmml/Test/samples/unittest_ref_ann_model.cpp
new file mode 100755 (executable)
index 0000000..643308c
--- /dev/null
+++ b/src/pmml/Test/samples/unittest_ref_ann_model.cpp
@@ -0,0 +1,67 @@
+#define ActivationFunction(sum) ( 1.0 / ( 1.0 + exp( -1.0 * sum )) )
+void myTestFunc(double *param, double *res)
+{
+  ////////////////////////////// 
+  //
+  // File used by unit test
+  // PMMLBasicsTest1::testExportNeuralNetworkCpp
+  //
+  ////////////////////////////// 
+
+  int nInput   = 8;
+  int nOutput   = 1;
+  int nHidden  = 1;
+  const int nNeurones  = 10;
+  double myTestFunc_act[nNeurones];
+
+  // --- Preprocessing of the inputs and outputs
+  double myTestFunc_minInput[] = {
+  0.099999, 25048.9, 89334.9, 89.5523, 1050, 
+  760.001, 1400.02, 10950, 
+  };
+  double myTestFunc_minOutput[] = {
+  77.8117,   };
+  double myTestFunc_maxInput[] = {
+  0.028899, 14419.8, 15180.8, 15.2866, 34.6793, 
+  34.6718, 161.826, 632.913, 
+  };
+  double myTestFunc_maxOutput[] = {
+  45.7061,   };
+
+  // --- Values of the weights
+  double myTestFunc_valW[] = {
+  -1.74548, 6.96551, -1.26357, 0.753663, 0.00165366, 
+  0.004725, 0.00996979, 0.178798, -0.180981, -0.173569, 
+  0.0855967, 
+  };
+  // --- Constants
+  int indNeurone = 0;
+  int CrtW;
+  double sum;
+
+  // --- Input Layers
+  for(int i = 0; i < nInput; i++) {
+     myTestFunc_act[indNeurone++] = ( param[i] - myTestFunc_minInput[i] ) / myTestFunc_maxInput[i];
+  }
+
+  // --- Hidden Layers
+  for (int member = 0; member < nHidden; member++) {
+     int CrtW = member * ( nInput + 2) + 2;
+     sum = myTestFunc_valW[CrtW++];
+     for (int source = 0; source < nInput; source++) {
+         sum += myTestFunc_act[source] * myTestFunc_valW[CrtW++];
+       }
+       myTestFunc_act[indNeurone++] = ActivationFunction(sum);
+  }
+
+  // --- Output
+  for (int member = 0; member < nOutput; member++) {
+    sum = myTestFunc_valW[0];
+    for (int source = 0; source < nHidden; source++) {
+      CrtW = source * ( nInput + 2) + 1;
+      sum += myTestFunc_act[nInput+source] * myTestFunc_valW[CrtW];
+    }
+    myTestFunc_act[indNeurone++] = sum;
+    res[member] = myTestFunc_minOutput[member] + myTestFunc_maxOutput[member] * sum;
+  }
+}