Refactoring package.mo into directories (#14)

* Refactoring package.mo into directories
* Refactoring code
  * Fixing white space
  * Adding html documentation to root package.mo

.github/CODEOWNERS

@@ -0,0 +1,5 @@
+# Each line is a file pattern followed by one or more owners.
+
+# License changes
+*license*     @bernhardbachmann
+*LICENSE*     @bernhardbachmann

.github/workflows/coverage.yml

@@ -35,7 +35,6 @@ jobs:
             omc
           libraries: |
             Modelica 4.0.0
-            ExternData 2.6.1
           omc-diff: true
 
       - name: openmodelica-library-testing

NeuralNetwork/Example_Models/Simple_Network/Model/Feedforward_Network.mo → Example_Models/Simple_Network/Model/Feedforward_Network.mo

@@ -5,14 +5,13 @@ model Feedforward_Network "Simulation of the simple network to demonstrate how t
     Placement(visible = true, transformation(origin = {-70, 50}, extent = {{-10, -10}, {10, 10}}, rotation = 0)));
   Modelica.Blocks.Sources.Constant const(k = 5) annotation(
     Placement(visible = true, transformation(origin = {-70, -50}, extent = {{-10, -10}, {10, 10}}, rotation = 0)));
-Network.Tree_Layer_Neural_Network tree_Layer_Neural_Network annotation(
+  Network.Tree_Layer_Neural_Network tree_Layer_Neural_Network annotation(
     Placement(visible = true, transformation(origin = {34, 0}, extent = {{-10, -10}, {10, 10}}, rotation = 0)));
 equation
   connect(ramp.y, mux.u[1]) annotation(
     Line(points = {{-58, 50}, {-40, 50}, {-40, 0}, {-18, 0}}, color = {0, 0, 127}));
   connect(const.y, mux.u[2]) annotation(
     Line(points = {{-58, -50}, {-40, -50}, {-40, 0}, {-18, 0}}, color = {0, 0, 127}));
-connect(mux.y, tree_Layer_Neural_Network.u) annotation(
+  connect(mux.y, tree_Layer_Neural_Network.u) annotation(
     Line(points = {{4, 0}, {22, 0}}, color = {0, 0, 127}, thickness = 0.5));
-protected
 end Feedforward_Network;

Example_Models/Simple_Network/Network/Tree_Layer_Neural_Network.mo

@@ -0,0 +1,44 @@
+block Tree_Layer_Neural_Network "This block models a very simple three layer Neural Network and should demonstrate the usage of the library"
+  // Parameter of Input Layer
+  parameter Real FirstLayer_bias[3] = {0, 0, 0} "Bias vector of the first layer";
+	parameter Real FirstLayer_weights[3, 2] = [1, 0; 0, 1; 0, 0] "Weight matrix of the first layer";
+  parameter Integer FirstLayer_numInputs = 2 "Specification of the input dimension of the neural network";
+  parameter Integer FirstLayer_numNeurons = 3 "Number of neurons in the first layer";
+  parameter Boolean FirstLayer_pca = false "Reducing the feature space based on a PCA";
+  parameter Boolean FirstLayer_scale = false "Scale the inputs with min-max normalization";
+  parameter Boolean FirstLayer_standardization = false "Scale the inputs with mu-sigma standardization";
+
+  // Parameter of First Hidden Layer
+  parameter Real SecondLayer_bias[4] = {0, 0, 0, 0} "Bias vector of the second layer";
+  parameter Real SecondLayer_weights[4, 3] = [1, 0, 0; 0, 1, 0; 0, 0, 0; 0, 0, 0] "Weight matrix of the second layer";
+  parameter Integer SecondLayer_numInputs = FirstLayer_numNeurons "Specification of the inputs of the second layer";
+  parameter Integer SecondLayer_numNeurons = 4 "Number of neurons in the second layer";
+  // Set the Activation Function
+
+  parameter Neural_Network.Activation_Function.Type_Activation_Function ActivationFunction = Neural_Network.Activation_Function.ReLu "Activation function of the Input and Hidden layers";
+
+  // Parameter of Output Layer
+  parameter Real ThirdLayer_bias[1] = {0} "Bias vector of the third layer";
+  parameter Real ThirdLayer_weights[1, 4] = [1, 1, 0, 0] "Weight matrix of the third layer";
+  parameter Integer ThirdLayer_numInputs = SecondLayer_numNeurons "Specification of the inputs of the third layer";
+  parameter Integer ThirdLayer_numNeurons = 1 "Number of neurons in the third layer";
+  parameter Boolean ThirdLayer_rescale = true "Rescale the outputs with min-max denormalization";
+  parameter Real ThirdLayer_max[1] = {5.0} "Maximum of the training targets";
+  parameter Real ThirdLayer_min[1] = {0.0} "Minimum of the training targets";
+  parameter Boolean ThirdLayer_destandardization = false "Destandardization of the outputs with mu-sigma";
+  parameter Real ThirdLayer_mean[1] = {0.0} "Default value";
+  parameter Real ThirdLayer_std[1] = {1.0} "Default value";
+
+  // Initilize the Layers of the neural network with the parameter that were defined above
+  Neural_Network.Layer.Input_Hidden_Layer First_Layer(bias = FirstLayer_bias, weights = FirstLayer_weights, NeuronActivation_Function = ActivationFunction, numInputs = FirstLayer_numInputs, numNeurons = FirstLayer_numNeurons, pca = FirstLayer_pca, scale = FirstLayer_scale, max = {1.0}, min = {0.0}, standardization = FirstLayer_standardization, mean = {0.0}, std = {1.0});
+  Neural_Network.Layer.Dense_Hidden_Layer Second_Layer(bias = SecondLayer_bias, weights = SecondLayer_weights, NeuronActivation_Function = ActivationFunction, numInputs = SecondLayer_numInputs, numNeurons = SecondLayer_numNeurons);
+  Neural_Network.Layer.Output_Hidden_Layer Third_Layer(bias = ThirdLayer_bias, weights = ThirdLayer_weights, numInputs = ThirdLayer_numInputs, numNeurons = ThirdLayer_numNeurons, rescale = ThirdLayer_rescale, max = ThirdLayer_max, min = ThirdLayer_min, destandardization = ThirdLayer_destandardization, mean = ThirdLayer_mean, std = ThirdLayer_std);
+
+  // Treat Neural Network as multiple Input - multiple Output Block
+  extends Modelica.Blocks.Interfaces.MIMO(final nin = FirstLayer_numInputs, final nout = ThirdLayer_numNeurons);
+equation
+  connect(u, First_Layer.u);
+  connect(First_Layer.y, Second_Layer.u);
+  connect(Second_Layer.y, Third_Layer.u);
+  connect(Third_Layer.y, y);
+end Tree_Layer_Neural_Network;

NeuralNetwork/Activation_Function/Type_Activation_Function.mo

@@ -0,0 +1,10 @@
+within NeuralNetwork.Activation_Function;
+
+type Type_Activation_Function "Temporary type of ActivationFunction with choices for menus (until enumerations are available)"
+  extends Integer;
+  annotation(
+    choices(
+      choice = NeuralNetwork.ActivationFunction.Type_Activation_Function.ReLu "ReLu activation function",
+      choice = NeuralNetwork.ActivationFunction.Type_Activation_Function.Sigmoid "Sigmoid activation function",
+      choise = NeuralNetwork.ActivationFunction.Type_Activation_Function.Hyperbolic_tangent "Tanh activation function"));
+end Type_Activation_Function;

NeuralNetwork/Activation_Function/package.mo

@@ -0,0 +1,7 @@
+within NeuralNetwork;
+
+package Activation_Function
+  constant Integer ReLu = 0;
+  constant Integer Sigmoid = 1;
+  constant Integer Hyperbolic_tangent = 2;
+end Activation_Function;

NeuralNetwork/Activation_Function/package.order

@@ -0,0 +1,4 @@
+ReLu
+Sigmoid
+Hyperbolic_tangent
+Type_Activation_Function