Changes in forward

src/cnn/creation.c

@@ -32,10 +32,10 @@ Network* create_network_lenet5(int dropout, int activation, int initialisation)
     Network* network = create_network(8, dropout, initialisation, 32, 1); 
     network->kernel[0]->activation = activation;  
     network->kernel[0]->linearisation = 0;
-    add_convolution(network, 6, 5, activation);
+    add_convolution(network, 1, 32, 6, 28, activation);
-    add_2d_average_pooling(network, 2);
+    add_2d_average_pooling(network, 28, 14);
-    add_convolution(network, 16, 5, activation);
+    add_convolution(network, 6, 14, 16, 10, activation);
-    add_2d_average_pooling(network, 2);
+    add_2d_average_pooling(network, 10, 5);
     add_dense_linearisation(network, 160, 120, activation);
     add_dense(network, 120, 84, activation);
     add_dense(network, 84, 10, SOFTMAX);
@@ -62,45 +62,38 @@ void create_a_line_input_layer(Network* network, int pos, int dim) {
     network->depth[pos] = 1;
 }
 
-void add_2d_average_pooling(Network* network, int kernel_size) {
+void add_2d_average_pooling(Network* network, int dim_input, int dim_ouput) {
     int n = network->size;
+    int k_pos = n-1;
     if (network->max_size == n) {
         printf("Impossible de rajouter une couche d'average pooling, le réseau est déjà plein\n");
         return;
     }
-    network->kernel[n]->cnn = NULL;
+    int kernel_size = dim_input/dim_ouput;
-    network->kernel[n]->nn = NULL;
+    if (dim_input%dim_ouput != 0) {
-    network->kernel[n]->activation = 100*kernel_size; // Ne contient pas de fonction d'activation
+        printf("Erreur de dimension dans l'average pooling\n");
+        return;
+    }
+    network->kernel[k_pos]->cnn = NULL;
+    network->kernel[k_pos]->nn = NULL;
+    network->kernel[k_pos]->activation = 100*kernel_size; // Ne contient pas de fonction d'activation
     create_a_cube_input_layer(network, n, network->depth[n-1], network->width[n-1]/2);
     network->size++;
 }
 
-void add_average_pooling_flatten(Network* network, int kernel_size) { // NEED TO BE VERIFIED
+void add_convolution(Network* network, int depth_input, int dim_input, int depth_output, int dim_output, int activation) {
-    int n = network->size;
-    if (network->max_size == n) {
-        printf("Impossible de rajouter une couche d'average pooling, le réseau est déjà plein\n");
-        return;
-    }
-    network->kernel[n]->cnn = NULL;
-    network->kernel[n]->nn = NULL;
-    network->kernel[n]->activation = 100*kernel_size; // Ne contient pas de fonction d'activation
-    int dim = (network->width[n-1]*network->width[n-1]*network->depth[n-1])/(kernel_size*kernel_size);
-    create_a_line_input_layer(network, n, dim);
-    network->size++;
-}
-
-void add_convolution(Network* network, int depth_output, int kernel_size, int activation) {
     int n = network->size;
+    int k_pos = n-1;
     if (network->max_size == n) {
         printf("Impossible de rajouter une couche de convolution, le réseau est déjà plein \n");
         return;
     }
-    int bias_size = network->width[n-1] - 2*(kernel_size/2);
+    int bias_size = dim_output;
-    int depth_input = network->depth[n-1];
+    int kernel_size = dim_input - dim_output +1;
-    network->kernel[n]->nn = NULL;
+    network->kernel[k_pos]->nn = NULL;
-    network->kernel[n]->activation = activation;
+    network->kernel[k_pos]->activation = activation;
-    network->kernel[n]->cnn = (Kernel_cnn*)malloc(sizeof(Kernel_cnn));
+    network->kernel[k_pos]->cnn = (Kernel_cnn*)malloc(sizeof(Kernel_cnn));
-    Kernel_cnn* cnn = network->kernel[n]->cnn;
+    Kernel_cnn* cnn = network->kernel[k_pos]->cnn;
 
     cnn->k_size = kernel_size;
     cnn->rows = depth_input;
@@ -132,23 +125,26 @@ void add_convolution(Network* network, int depth_output, int kernel_size, int ac
     create_a_cube_input_layer(network, n, depth_output, bias_size);
     int n_int = network->width[n-1]*network->width[n-1]*network->depth[n-1];
     int n_out = network->width[n]*network->width[n]*network->depth[n];
+    /* Not currently used 
     initialisation_3d_matrix(network->initialisation, cnn->bias, depth_output, kernel_size, kernel_size, n_int+n_out);
     initialisation_3d_matrix(ZERO, cnn->d_bias, depth_output, kernel_size, kernel_size, n_int+n_out);
     initialisation_4d_matrix(network->initialisation, cnn->w, depth_input, depth_output, kernel_size, kernel_size, n_int+n_out);
     initialisation_4d_matrix(ZERO, cnn->d_w, depth_input, depth_output, kernel_size, kernel_size, n_int+n_out);
+    */
     network->size++;
 }
 
 void add_dense(Network* network, int input_units, int output_units, int activation) {
     int n = network->size;
+    int k_pos = n-1;
     if (network->max_size == n) {
         printf("Impossible de rajouter une couche dense, le réseau est déjà plein\n");
         return;
     }
-    network->kernel[n]->cnn = NULL;
+    network->kernel[k_pos]->cnn = NULL;
-    network->kernel[n]->nn = (Kernel_nn*)malloc(sizeof(Kernel_nn));
+    network->kernel[k_pos]->nn = (Kernel_nn*)malloc(sizeof(Kernel_nn));
-    Kernel_nn* nn = network->kernel[n]->nn;
+    Kernel_nn* nn = network->kernel[k_pos]->nn;
-    network->kernel[n]->activation = activation;
+    network->kernel[k_pos]->activation = activation;
     nn->input_units = input_units;
     nn->output_units = output_units;
     nn->bias = (float*)malloc(sizeof(float)*output_units);
@@ -159,11 +155,12 @@ void add_dense(Network* network, int input_units, int output_units, int activati
         nn->weights[i] = (float*)malloc(sizeof(float)*output_units);
         nn->d_weights[i] = (float*)malloc(sizeof(float)*output_units);
     }
+    /* Not currently used
     initialisation_1d_matrix(network->initialisation, nn->bias, output_units, output_units+input_units);
     initialisation_1d_matrix(ZERO, nn->d_bias, output_units, output_units+input_units);
     initialisation_2d_matrix(network->initialisation, nn->weights, input_units, output_units, output_units+input_units);
     initialisation_2d_matrix(ZERO, nn->d_weights, input_units, output_units, output_units+input_units);
-    create_a_line_input_layer(network, n, output_units);
+    create_a_line_input_layer(network, n, output_units); */
     network->size++;
 }
 
@@ -171,14 +168,15 @@ void add_dense_linearisation(Network* network, int input_units, int output_units
     // Can replace input_units by a research of this dim
 
     int n = network->size;
+    int k_pos = n-1;
     if (network->max_size == n) {
         printf("Impossible de rajouter une couche dense, le réseau est déjà plein\n");
         return;
     }
-    network->kernel[n]->cnn = NULL;
+    network->kernel[k_pos]->cnn = NULL;
-    network->kernel[n]->nn = (Kernel_nn*)malloc(sizeof(Kernel_nn));
+    network->kernel[k_pos]->nn = (Kernel_nn*)malloc(sizeof(Kernel_nn));
-    Kernel_nn* nn = network->kernel[n]->nn;
+    Kernel_nn* nn = network->kernel[k_pos]->nn;
-    network->kernel[n]->activation = activation;
+    network->kernel[k_pos]->activation = activation;
     nn->input_units = input_units;
     nn->output_units = output_units;
     
@@ -190,10 +188,11 @@ void add_dense_linearisation(Network* network, int input_units, int output_units
         nn->weights[i] = (float*)malloc(sizeof(float)*output_units);
         nn->d_weights[i] = (float*)malloc(sizeof(float)*output_units);
     }
+    /* Not currently used
     initialisation_1d_matrix(network->initialisation, nn->bias, output_units, output_units+input_units);
     initialisation_1d_matrix(ZERO, nn->d_bias, output_units, output_units+input_units);
     initialisation_2d_matrix(network->initialisation, nn->weights, input_units, output_units, output_units+input_units);
-    initialisation_2d_matrix(ZERO, nn->d_weights, input_units, output_units, output_units+input_units);
+    initialisation_2d_matrix(ZERO, nn->d_weights, input_units, output_units, output_units+input_units); */
     create_a_line_input_layer(network, n, output_units);
 
     network->size++;

src/cnn/function.c

@@ -72,20 +72,38 @@ void apply_function_input(float (*f)(float), float*** input, int depth, int rows
     }
 }
 
-void choose_apply_function_input(int activation, float*** input, int depth, int rows, int columns) {
+void choose_apply_function_matrix(int activation, float*** input, int depth, int dim) {
     if (activation == RELU) {
-        apply_function_input(relu, input, depth, rows, columns);
+        apply_function_input(relu, input, depth, dim, dim);
     }
     else if (activation == SIGMOID) {
-        apply_function_input(sigmoid, input, depth, rows, columns);
+        apply_function_input(sigmoid, input, depth, dim, dim);
     }
     else if (activation == SOFTMAX) {
-        apply_softmax_input(input, depth, rows, columns);
+        apply_softmax_input(input, depth, dim, dim);
     }
     else if (activation == TANH) {
-        apply_function_input(tanh_, input, depth, rows, columns);
+        apply_function_input(tanh_, input, depth, dim, dim);
     }
     else {
-        printf("Erreur, fonction d'activation inconnue: %d\n", activation);
+        printf("Erreur, fonction d'activation inconnue (choose_apply_function_matrix): %d\n", activation);
+    }
+}
+
+void choose_apply_function_vector(int activation, float*** input, int dim) {
+    if (activation == RELU) {
+        apply_function_input(relu, input, 1, 1, dim);
+    }
+    else if (activation == SIGMOID) {
+        apply_function_input(sigmoid, input, 1, 1, dim);
+    }
+    else if (activation == SOFTMAX) {
+        apply_softmax_input(input, 1, 1, dim);
+    }
+    else if (activation == TANH) {
+        apply_function_input(tanh_, input, 1, 1, dim);
+    }
+    else {
+        printf("Erreur, fonction d'activation inconnue (choose_apply_function_vector): %d\n", activation);
     }
 }

src/cnn/include/creation.h

@@ -27,17 +27,12 @@ void create_a_line_input_layer(Network* network, int pos, int dim);
 /*
 * Ajoute au réseau une couche d'average pooling valide de dimension dim*dim
 */
-void add_2d_average_pooling(Network* network, int kernel_size);
+void add_2d_average_pooling(Network* network, int dim_input, int dim_ouput);
-
-/*
-* Ajoute au réseau une couche d'average pooling valide de dimension dim*dim qui aplatit
-*/
-void add_average_pooling_flatten(Network* network, int kernel_size);
 
 /*
 * Ajoute au réseau une couche de convolution dim*dim et initialise les kernels
 */
-void add_convolution(Network* network, int nb_filter, int kernel_size, int activation);
+void add_convolution(Network* network, int depth_input, int dim_input, int depth_output, int dim_output, int activation);
 
 /*
 * Ajoute au réseau une couche dense et initialise les poids et les biais

src/cnn/include/function.h

@@ -35,9 +35,14 @@ void apply_softmax_input(float ***input, int depth, int rows, int columns);
 void apply_function_input(float (*f)(float), float*** input, int depth, int rows, int columns);
 
 /*
-* Redirige vers la fonction à appliquer sur ????
+* Redirige vers la fonction à appliquer sur une matrice
 */
-void choose_apply_function_input(int activation, float*** input, int depth, int rows, int columns);
+void choose_apply_function_matrix(int activation, float*** input, int depth, int dim);
+
+/*
+* Redirige vers la fonction à appliquer sur un vecteur
+*/
+void choose_apply_function_vector(int activation, float*** input, int dim);
 
 
 #endif

src/cnn/include/make.h

@@ -13,11 +13,6 @@ void make_convolution(float*** input, Kernel_cnn* kernel, float*** output, int o
 */
 void make_average_pooling(float*** input, float*** output, int size, int output_depth, int output_dim);
 
-/*
-* Effectue un average pooling avec stride=size et aplatissement
-*/
-void make_average_pooling_flattened(float*** input, float* output, int size, int input_depth, int input_dim);
-
 /*
 * Effecute une full connection
 */

src/cnn/main.c

@@ -36,44 +36,40 @@ void forward_propagation(Network* network) {
     int n = network->size;
     float*** input;
     float*** output;
-    Kernel* k_i_1;
     Kernel* k_i;
     for (int i=0; i < n-1; i++) {
-        k_i_1 = network->kernel[i+1];
         k_i = network->kernel[i];
+        printf("\n i -> %d :: %d %d \n", i, k_i->cnn==NULL, k_i->nn==NULL);
         input_width = network->width[i];
         input_depth = network->depth[i];
         output_width = network->width[i+1];
         output_depth = network->depth[i+1];
-        activation = network->kernel[i]->activation;
+        activation = k_i->activation;
         input = network->input[i];
         output = network->input[i+1];
 
-        if (k_i_1->nn==NULL && k_i_1->cnn!=NULL) { //CNN
+        if (k_i->cnn!=NULL) { //CNN
-            printf("Convolution of cnn: %dx%d -> %dx%d\n", input_depth, input_width, output_depth, output_width);
+            printf("Convolution of cnn: %dx%dx%d -> %dx%dx%d\n", input_depth, input_width, input_width, output_depth, output_width, output_width);
-            make_convolution(input, k_i_1->cnn, output, output_width);
+            make_convolution(input, k_i->cnn, output, output_width);
-            choose_apply_function_input(activation, output, output_depth, output_width, output_width);
+            choose_apply_function_matrix(activation, output, output_depth, output_width);
         }
-        else if (k_i_1->nn!=NULL && k_i_1->cnn==NULL) { //NN
+        else if (k_i->nn!=NULL) { //NN
-            printf("Densification of nn\n");
+            printf("Densification of nn: %dx%dx%d -> %dx%dx%d\n", input_depth, input_width, input_width, output_depth, output_width, output_width);
             // Checked if it is a nn which linearise
             make_fully_connected(network->input[i][0][0], network->kernel[i]->nn, network->input[i+1][0][0], input_width, output_width);
-            choose_apply_function_input(activation, output, 1, 1, output_width);
+            choose_apply_function_vector(activation, output, output_width);
         }
-        else { //Pooling (Vérifier dedans) ??
+        else { //Pooling
             if (n-2==i) {
                 printf("Le réseau ne peut pas finir par une pooling layer");
                 return;
             }
             if (1==1) { // Pooling sur une matrice
-                printf("Average pooling\n");
+                printf("Average pooling: %dx%dx%d -> %dx%dx%d\n", input_depth, input_width, input_width, output_depth, output_width, output_width);
                 make_average_pooling(input, output, activation/100, output_depth, output_width);
             }
-            else if (1==0) { // Pooling sur un vecteur
+            else { // Pooling sur un vecteur
-                printf("Error: Not implemented: forward: %d\n", i);
+                printf("Erreur: le pooling ne se fait que sur une matrice \n");
-            }
-            else {
-                printf("Erreur: forward_propagation: %d -> %d %d\n", i, k_i_1->nn==NULL, k_i_1->cnn);
                 return;
             }
         }
@@ -89,7 +85,7 @@ void backward_propagation(Network* network, float wanted_number) { // TODO
         if (i==n) {
             if (network->kernel[i]->activation == SOFTMAX) {
                 int l2 = network->width[i]; // Taille de la dernière couche
-                int l1 = network->width[i-1];
+                //int l1 = network->width[i-1];
                 for (int j=0; j < l2; j++) {
 
                 }

src/cnn/make.c

@@ -4,12 +4,9 @@
 #include "include/make.h"
 
 void make_convolution(float*** input, Kernel_cnn* kernel, float*** output, int output_dim) {
-    // TODO, MISS CONDITIONS ON THE CONVOLUTION
     printf_warning("Appel de make_convolution, incomplet\n");
     float f;
     int n = kernel->k_size;
-    printf("Convolution output: %dx%dx%d, %dx%dx%d\n", kernel->columns, output_dim, output_dim, kernel->rows, n, n);
-    printf("BIS %d %d \n", kernel->columns, kernel->k_size);
     for (int i=0; i < kernel->columns; i++) {
         for (int j=0; j < output_dim; j++) {
             for (int k=0; k < output_dim; k++) {
@@ -47,31 +44,6 @@ void make_average_pooling(float*** input, float*** output, int size, int output_
     }
 }
 
-void make_average_pooling_flattened(float*** input, float* output, int size, int input_depth, int input_dim) {
-    if ((input_depth*input_dim*input_dim) % (size*size) != 0) {
-        printf_error("Deux layers non compatibles avec un average pooling flattened");
-        return;
-    }
-    float average;
-    int n = size*size;
-    int cpt = 0;
-    int output_dim = input_dim - 2*(size/2);
-    for (int i=0; i < input_depth; i++) {
-        for (int j=0; j < output_dim; j++) {
-            for (int k=0; k < output_dim; k++) {
-                average = 0.;
-                for (int a=0; a < size; a++) {
-                    for (int b=0; b < size; b++) {
-                        average += input[i][2*j +a][2*k +b];
-                    }
-                }
-                output[cpt] = average;
-                cpt++;
-            }
-        }
-    }
-}
-
 void make_fully_connected(float* input, Kernel_nn* kernel, float* output, int size_input, int size_output) {
     float f;
     for (int i=0; i < size_output; i++) {