Fixed MultiNEAT Neural Network bug

the NN needs to be activated as many times as the network is deep

cpprevolve/revolve/brains/controller/DifferentialCPG.cpp

@@ -94,6 +94,7 @@ DifferentialCPG::DifferentialCPG(
     // build the NN according to the genome
     NEAT::NeuralNetwork net;
     gen.BuildPhenotype(net);
+    unsigned int net_depth = net.CalculateNetworkDepth();
 
     // get weights for each connection
     // assuming that connections are distinct for each direction
@@ -109,9 +110,12 @@ DifferentialCPG::DifferentialCPG(
         inputs[3] = 1;
         std::tie(inputs[4], inputs[5], inputs[6]) = motor.first;
         inputs[7] = -1;
+        inputs[8] = 1;
 
+        net.Flush();
         net.Input(inputs);
-        net.Activate();
+        for (int i=0; i<net_depth; i++)
+            net.Activate();
         double weight = net.Output()[0];
 #ifdef DifferentialCPG_PRINT_INFO
         std::cout << "Creating weight ["
@@ -127,8 +131,12 @@ DifferentialCPG::DifferentialCPG(
         int k = con.second;
         // convert tuple to vector
         std::tie(inputs[0], inputs[1], inputs[2], inputs[3], inputs[4], inputs[5], inputs[6], inputs[7]) = con.first;
+        inputs[8] = 1;
+
+        net.Flush();
         net.Input(inputs);
-        net.Activate();
+        for (int i=0; i<net_depth; i++)
+            net.Activate();
         double weight = net.Output()[0];
 #ifdef DifferentialCPG_PRINT_INFO
         std::cout << "Creating weight ["