rawalcher/c-net
Archived
1
0
Fork 0
Code Issues 5 Pull requests Projects Releases 2 Packages Wiki Activity Actions

fixed pointer stuff

Browse source
This commit is contained in:
Thomas 2023-09-22 09:28:01 +02:00
parent 5774ddc116
commit 86ac3e855c
1 changed files with 28 additions and 23 deletions
Download patch file Download diff file Expand all files Collapse all files

51
neuronal_network.c
View file

@ -12,7 +12,7 @@ double square(double input);
double loss_function(Matrix* output_matrix, int image_label); double loss_function(Matrix* output_matrix, int image_label);
void backPropagation(double learning_rate, Matrix* weights, Matrix* biases, Matrix* current_layer_activation, Matrix* previous_layer_activation, Matrix* sigma_old); Matrix * backPropagation(double learning_rate, Matrix* weights, Matrix* biases, Matrix* current_layer_activation, Matrix* previous_layer_activation, Matrix* sigma_old);
Neural_Network* new_network(int input_size, int hidden_size, int output_size, double learning_rate){ Neural_Network* new_network(int input_size, int hidden_size, int output_size, double learning_rate){
Neural_Network *network = malloc(sizeof(Neural_Network)); Neural_Network *network = malloc(sizeof(Neural_Network));
@ -226,20 +226,20 @@ void train_network(Neural_Network* network, Image *image, int label) {
Matrix* final_outputs = apply(sigmoid, final_add); Matrix* final_outputs = apply(sigmoid, final_add);
// begin backpropagation // begin backpropagation
Matrix* sigma = matrix_create(final_outputs->rows, 1); Matrix* sigma1 = matrix_create(final_outputs->rows, 1);
matrix_fill(sigma, 1); matrix_fill(sigma1, 1);
Matrix* temp1 = subtract(sigma, final_outputs); Matrix* temp1 = subtract(sigma1, final_outputs);
Matrix* temp2 = multiply(temp1, final_outputs); // * soll-ist Matrix* temp2 = multiply(temp1, final_outputs); // * soll-ist
Matrix* temp3 = matrix_create(final_outputs->rows, final_outputs->columns); Matrix* temp3 = matrix_create(final_outputs->rows, final_outputs->columns);
matrix_fill(temp3, 0); matrix_fill(temp3, 0);
temp3->numbers[label][0] = 1; temp3->numbers[label][0] = 1;
Matrix* temp4 = subtract(temp3, final_outputs); Matrix* temp4 = subtract(temp3, final_outputs);
sigma = multiply(temp2, temp4); sigma1 = multiply(temp2, temp4);
Matrix* temp5 = transpose(h3_outputs); Matrix* temp5 = transpose(h3_outputs);
Matrix* temp6 = dot(sigma, temp5); Matrix* temp6 = dot(sigma1, temp5);
Matrix* weights_delta = scale(temp6, network->learning_rate); Matrix* weights_delta = scale(temp6, network->learning_rate);
Matrix* bias_delta = scale(sigma, network->learning_rate); Matrix* bias_delta = scale(sigma1, network->learning_rate);
Matrix* temp7 = add(weights_delta, network->weights_output); Matrix* temp7 = add(weights_delta, network->weights_output);
matrix_free(network->weights_output); matrix_free(network->weights_output);
@ -250,9 +250,9 @@ void train_network(Neural_Network* network, Image *image, int label) {
network->bias_output = temp8; network->bias_output = temp8;
// other levels // other levels
backPropagation(network->learning_rate, network->weights_3, network->bias_3, h3_outputs, h2_outputs, sigma); Matrix* sigma2 = backPropagation(network->learning_rate, network->weights_3, network->bias_3, h3_outputs, h2_outputs, sigma1);
backPropagation(network->learning_rate, network->weights_2, network->bias_2, h2_outputs, h1_outputs, sigma); Matrix* sigma3 = backPropagation(network->learning_rate, network->weights_2, network->bias_2, h2_outputs, h1_outputs, sigma2);
backPropagation(network->learning_rate, network->weights_1, network->bias_1, h1_outputs, input, sigma); Matrix* sigma4 = backPropagation(network->learning_rate, network->weights_1, network->bias_1, h1_outputs, input, sigma3);
matrix_free(input); matrix_free(input);
@ -275,7 +275,11 @@ void train_network(Neural_Network* network, Image *image, int label) {
matrix_free(weights_delta); matrix_free(weights_delta);
matrix_free(bias_delta); matrix_free(bias_delta);
matrix_free(sigma); matrix_free(sigma1);
matrix_free(sigma2);
matrix_free(sigma3);
matrix_free(sigma4);
matrix_free(temp1); matrix_free(temp1);
matrix_free(temp2); matrix_free(temp2);
@ -287,7 +291,7 @@ void train_network(Neural_Network* network, Image *image, int label) {
matrix_free(temp8); matrix_free(temp8);
} }
void backPropagation(double learning_rate, Matrix* weights, Matrix* biases, Matrix* current_layer_activation, Matrix* previous_layer_activation, Matrix* sigma_old) { Matrix * backPropagation(double learning_rate, Matrix* weights, Matrix* biases, Matrix* current_layer_activation, Matrix* previous_layer_activation, Matrix* sigma_old) {
Matrix* sigma_new = matrix_create(current_layer_activation->rows, 1); Matrix* sigma_new = matrix_create(current_layer_activation->rows, 1);
matrix_fill(sigma_new, 1); matrix_fill(sigma_new, 1);
@ -311,19 +315,18 @@ void backPropagation(double learning_rate, Matrix* weights, Matrix* biases, Matr
Matrix* bias_delta = scale(sigma_new, learning_rate); Matrix* bias_delta = scale(sigma_new, learning_rate);
Matrix* temp5 = add(weights_delta, weights); Matrix* temp5 = add(weights_delta, weights);
free(weights->numbers); for (int i = 0; i < weights->rows; ++i) {
weights->numbers = temp5->numbers; for (int j = 0; j < weights->columns; ++j) {
weights->numbers[i][j] = temp5->numbers[i][j];
}
}
Matrix* temp6 = add(bias_delta, biases); Matrix* temp6 = add(bias_delta, biases);
free(biases->numbers); for (int i = 0; i < biases->rows; ++i) {
biases->numbers = temp6->numbers; for (int j = 0; j < biases->columns; ++j) {
biases->numbers[i][j] = temp6->numbers[i][j];
sigma_old->rows = sigma_new->rows; }
sigma_old->columns = sigma_new->columns; }
free(sigma_old->numbers);
sigma_old->numbers = sigma_new->numbers;
free(sigma_new);
matrix_free(temp1); matrix_free(temp1);
matrix_free(temp2); matrix_free(temp2);
@ -333,6 +336,8 @@ void backPropagation(double learning_rate, Matrix* weights, Matrix* biases, Matr
matrix_free(temp6); matrix_free(temp6);
matrix_free(weights_delta); matrix_free(weights_delta);
matrix_free(bias_delta); matrix_free(bias_delta);
return sigma_new;
} }