update neuralNetwork

imageInput.c

@@ -33,6 +33,7 @@ GrayScaleImageSeries *readImages(const char *path)
         return NULL;
     }
     //liest die Anzahl der Bilder aus
+    series->count = 0;
     fread(&series->count, sizeof(unsigned short),1, data);
     series->images = malloc(series->count * sizeof(GrayScaleImage));
     if (series->images == NULL){

imageInput.h

@@ -12,7 +12,7 @@ typedef struct
 
 typedef struct
 {
-    GrayScaleImage *images; //in sich verschachtelte Struktur
+    GrayScaleImage *images;
     unsigned char *labels;
     unsigned int count;
 } GrayScaleImageSeries;

imageInputTests.c

@@ -123,7 +123,8 @@ void setUp(void) {
     // Falls notwendig, kann hier Vorbereitungsarbeit gemacht werden
 }
 
-void tearDown(void) {
+void tearDown(void) 
+{
     // Hier kann Bereinigungsarbeit nach jedem Test durchgeführt werden
 }
 

matrix.c

@@ -2,20 +2,22 @@
 #include <string.h>
 #include "matrix.h"
 
+// TODO Matrix-Funktionen implementieren
+
 Matrix createMatrix(unsigned int rows, unsigned int cols)
 {
-    Matrix m = {NULL, 0, 0};
+    Matrix matrix = {NULL, 0, 0};
 
     if (rows == 0 || cols == 0)
-        return m;
+        return matrix;   //gibt leere Matrix zurück
 
-    m.buffer = (MatrixType *)calloc(rows * cols, sizeof(MatrixType));
+    matrix.buffer = (MatrixType *)calloc(rows * cols, sizeof(MatrixType));
-    if (m.buffer == NULL)
+    if (matrix.buffer == NULL) //auf verfügbaren Speicherplatz prüfen
-        return m;
+        return matrix;
 
-    m.rows = rows;
+    matrix.rows = rows;  
-    m.cols = cols;
+    matrix.cols = cols;
-    return m;
+    return matrix;   //Matrix zurückgeben
 }
 
 void clearMatrix(Matrix *matrix)
@@ -23,18 +25,18 @@ void clearMatrix(Matrix *matrix)
 
     if (matrix != NULL)
     {
-        free(matrix->buffer);
+        free(matrix->buffer);   //Speicherplatz bereinigen
-        matrix->buffer = NULL;
+        matrix->buffer = NULL;  //Werte auf 0 setzen
         matrix->rows = 0;
         matrix->cols = 0;
     }
 }
 
 void setMatrixAt(MatrixType value, Matrix matrix, unsigned int rowIdx, unsigned int colIdx) 
-// Matrix matrix zu Matrix *matrix, empfehlung
 {
-    if (rowIdx < matrix.rows && colIdx < matrix.cols && matrix.buffer != NULL)
+    if (rowIdx < matrix.rows && colIdx < matrix.cols && matrix.buffer != NULL) //Prüft ob Zugriff möglich
         matrix.buffer[rowIdx * matrix.cols + colIdx] = value; 
+        //schreibt 2D element in 1D Liste: Element_Reihe*Matrix_Spalten + Element_Spalte
 }
 
 MatrixType getMatrixAt(const Matrix matrix, unsigned int rowIdx, unsigned int colIdx)
@@ -45,11 +47,10 @@ MatrixType getMatrixAt(const Matrix matrix, unsigned int rowIdx, unsigned int co
 }
 
 // TODO: Funktionen implementieren
-
 Matrix add(const Matrix matrix1, const Matrix matrix2)
 {
-    //  check, equal rows
+    //  immer Probe, gleiche Zeilen der Matrizen
-    // "Elementweise Addition": test, if two matrix has exact size
+    // "Elementweise Addition": Probe, ob matrix gleiche größe hat
     if (matrix1.rows == matrix2.rows && matrix1.cols == matrix2.cols)
     {
         Matrix result_add = createMatrix(matrix1.rows, matrix1.cols);
@@ -64,7 +65,7 @@ Matrix add(const Matrix matrix1, const Matrix matrix2)
         }
         return result_add;
     }
-    // "Broadcasting": matrix1 has 1 collum
+    // "Broadcasting": matrix1 hat 1 Spalte
     if (matrix1.rows == matrix2.rows && matrix1.cols == 1)
     {
         Matrix result_add = createMatrix(matrix1.rows, matrix2.cols);
@@ -78,7 +79,7 @@ Matrix add(const Matrix matrix1, const Matrix matrix2)
         }
         return result_add;
     }
-    // "Broadcasting": matrix2 has 1 collum
+    // "Broadcasting": matrix2 hat 1 Spalte
     if (matrix1.rows == matrix2.rows && matrix2.cols == 1)
     {
         Matrix result_add = createMatrix(matrix1.rows, matrix1.cols);
@@ -98,13 +99,14 @@ Matrix add(const Matrix matrix1, const Matrix matrix2)
 
 Matrix multiply(const Matrix matrix1, const Matrix matrix2)
 {
-    // Needed: rows/Zeilen, collums/Spalten
     MatrixType buffer_add;
-    // Probe ob Spalten1 = Zeilen2
+    
-    if (matrix1.cols != matrix2.rows)
+    if (!matrix1.buffer || !matrix2.buffer) // Probe ob leere Matrize vorliegt 
+        return createMatrix(0, 0);
+    if (matrix1.cols != matrix2.rows) // Probe ob Spalten1 = Zeilen2
         return createMatrix(0, 0);
 
-    Matrix result_mul = createMatrix(matrix1.rows, matrix2.cols); // ""
+    Matrix result_mul = createMatrix(matrix1.rows, matrix2.cols);
 
     for (unsigned int index = 0; index < matrix1.rows; index++)
     {
@@ -113,11 +115,9 @@ Matrix multiply(const Matrix matrix1, const Matrix matrix2)
             buffer_add = 0;
             for (unsigned int skalar = 0; skalar < matrix1.cols; skalar++)
             {
-                // buffer_add += matrix1[index][skalar]*matrix2[skalar][shift];
                 buffer_add += getMatrixAt(matrix1, index, skalar) * getMatrixAt(matrix2, skalar, shift);
             }
-            // matrix_mul[index][shift] = buffer_add;
+            setMatrixAt(buffer_add, result_mul, index, shift);
-            setMatrixAt(buffer_add, result_mul, index, shift); // result als Pointer, also mit &result
         }
     }
     return result_mul;

matrix.h

@@ -7,7 +7,7 @@ typedef float MatrixType;
 
 // TODO Matrixtyp definieren
 typedef struct {
-    MatrixType *matrix;
+    MatrixType *buffer;
     unsigned int rows;
     unsigned int cols;
 } Matrix;

neuralNetworkTests.c

@@ -5,10 +5,49 @@
 #include "unity.h"
 #include "neuralNetwork.h"
 
-
 static void prepareNeuralNetworkFile(const char *path, const NeuralNetwork nn)
 {
-    // TODO
+    FILE *file = fopen(path, "wb");
+    if (!file)
+        return;
+
+    const char *tag = "__info2_neural_network_file_format__";
+    fwrite(tag, 1, strlen(tag), file);
+
+    // Überprüfung, ob es Layer gibt
+    if (nn.numberOfLayers == 0)
+    {
+        fclose(file);
+        return;
+    }
+
+    // Schreibe die Eingabe- und Ausgabegrößen des Netzwerks
+    int input = nn.layers[0].weights.cols;
+    int output = nn.layers[0].weights.rows;
+
+    fwrite(&input, sizeof(int), 1, file);
+    fwrite(&output, sizeof(int), 1, file);
+
+    // Schreibe die Layer-Daten
+    for (int i = 0; i < nn.numberOfLayers; i++)
+    {
+        const Layer *layer = &nn.layers[i];
+        int out = layer->weights.rows;
+        int in = layer->weights.cols;
+
+        fwrite(layer->weights.buffer, sizeof(MatrixType), out * in, file);
+
+        fwrite(layer->biases.buffer, sizeof(MatrixType), out * 1, file);
+
+        if (i + 1 < nn.numberOfLayers)
+        {
+            int nextOut = nn.layers[i + 1].weights.rows;
+            fwrite(&nextOut, sizeof(int), 1, file);
+        }
+    }
+    fclose(file);
+
+   
 }
 
 void test_loadModelReturnsCorrectNumberOfLayers(void)
@@ -205,8 +244,8 @@ void test_predictReturnsCorrectLabels(void)
     Matrix biases1 = {.buffer = biasBuffer1, .rows = 2, .cols = 1};
     Matrix biases2 = {.buffer = biasBuffer2, .rows = 3, .cols = 1};
     Matrix biases3 = {.buffer = biasBuffer3, .rows = 5, .cols = 1};
-    Layer layers[] = {{.weights=weights1, .biases=biases1, .activation=someActivation}, \
+    Layer layers[] = {{.weights = weights1, .biases = biases1, .activation = someActivation},
-                      {.weights=weights2, .biases=biases2, .activation=someActivation}, \
+                      {.weights = weights2, .biases = biases2, .activation = someActivation},
                       {.weights = weights3, .biases = biases3, .activation = someActivation}};
     NeuralNetwork netUnderTest = {.layers = layers, .numberOfLayers = 3};
     unsigned char *predictedLabels = predict(netUnderTest, inputImages, 2);
@@ -216,11 +255,13 @@ void test_predictReturnsCorrectLabels(void)
     free(predictedLabels);
 }
 
-void setUp(void) {
+void setUp(void)
+{
     // Falls notwendig, kann hier Vorbereitungsarbeit gemacht werden
 }
 
-void tearDown(void) {
+void tearDown(void)
+{
     // Hier kann Bereinigungsarbeit nach jedem Test durchgeführt werden
 }