File size: 9,581 Bytes
00e6582
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
#include "neural_network.h"


#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <time.h>

// --------------------------------------------------------
// 1. Structures
// --------------------------------------------------------

// --------------------------------------------------------
// 2. Math & Activations
// --------------------------------------------------------

// Generates a random double between -0.5 and 0.5
double random_weight() {
    return ((double)rand() / (double)RAND_MAX) - 0.5;
}

// Sigmoid activation function
double sigmoid(double x) {
    return 1.0 / (1.0 + exp(-x));
}

// Derivative of sigmoid (assuming x is already activated)
double sigmoid_derivative(double x) {
    return x * (1.0 - x);
}

// --------------------------------------------------------
// 3. Initialization & Memory Management
// --------------------------------------------------------

// Initialize a single NNLayer
void init_NNLayer(NNLayer *NNLayer, int num_inputs, int num_neurons) {
    NNLayer->num_inputs = num_inputs;
    NNLayer->num_neurons = num_neurons;

    NNLayer->weights = (double*)malloc(num_inputs * num_neurons * sizeof(double));
    NNLayer->biases = (double*)malloc(num_neurons * sizeof(double));
    NNLayer->outputs = (double*)malloc(num_neurons * sizeof(double));
    NNLayer->deltas = (double*)malloc(num_neurons * sizeof(double));

    for (int i = 0; i < num_inputs * num_neurons; i++) {
        NNLayer->weights[i] = random_weight();
    }
    for (int i = 0; i < num_neurons; i++) {
        NNLayer->biases[i] = random_weight();
        NNLayer->outputs[i] = 0.0;
        NNLayer->deltas[i] = 0.0;
    }
}

// Create the network
Network* create_network(int num_inputs, int num_hidden_NNLayers, int hidden_size, int num_outputs) {
    Network *net = (Network*)malloc(sizeof(Network));
    net->num_inputs = num_inputs;
    net->num_NNLayers = num_hidden_NNLayers + 1; // Hidden + 1 Output NNLayer
    net->NNLayers = (NNLayer*)malloc(net->num_NNLayers * sizeof(NNLayer));

    // First hidden NNLayer
    init_NNLayer(&net->NNLayers[0], num_inputs, hidden_size);

    // Additional hidden NNLayers
    for (int i = 1; i < num_hidden_NNLayers; i++) {
        init_NNLayer(&net->NNLayers[i], hidden_size, hidden_size);
    }

    // Output NNLayer
    init_NNLayer(&net->NNLayers[net->num_NNLayers - 1], hidden_size, num_outputs);

    return net;
}

// Free memory
void free_network(Network *net) {
    for (int i = 0; i < net->num_NNLayers; i++) {
        free(net->NNLayers[i].weights);
        free(net->NNLayers[i].biases);
        free(net->NNLayers[i].outputs);
        free(net->NNLayers[i].deltas);
    }
    free(net->NNLayers);
    free(net);
}

// --------------------------------------------------------
// 4. Feedforward & Backpropagation
// --------------------------------------------------------

// Forward pass
double* feed_forward(Network *net, double *inputs) {
    for (int l = 0; l < net->num_NNLayers; l++) {
        NNLayer *NNLayer = &net->NNLayers[l];
        double *NNLayer_inputs = (l == 0) ? inputs : net->NNLayers[l-1].outputs;

        for (int j = 0; j < NNLayer->num_neurons; j++) {
            double activation = NNLayer->biases[j];
            for (int i = 0; i < NNLayer->num_inputs; i++) {
                // index = neuron_index * num_inputs + input_index
                activation += NNLayer_inputs[i] * NNLayer->weights[j * NNLayer->num_inputs + i];
            }
            NNLayer->outputs[j] = sigmoid(activation);
        }
    }
    return net->NNLayers[net->num_NNLayers - 1].outputs; // Return final NNLayer outputs
}

// Train a single sample
void backpropagate(Network *net, double *inputs, double *targets, double learning_rate) {
    // 1. Perform forward pass
    feed_forward(net, inputs);

    // 2. Calculate Output NNLayer Deltas (Error * Derivative)
    NNLayer *output_NNLayer = &net->NNLayers[net->num_NNLayers - 1];
    for (int j = 0; j < output_NNLayer->num_neurons; j++) {
        double output = output_NNLayer->outputs[j];
        double error = targets[j] - output;
        output_NNLayer->deltas[j] = error * sigmoid_derivative(output);
    }

    // 3. Calculate Hidden NNLayer Deltas (Propagate backwards)
    for (int l = net->num_NNLayers - 2; l >= 0; l--) {
        NNLayer *current_NNLayer = &net->NNLayers[l];
        NNLayer *next_NNLayer = &net->NNLayers[l + 1];

        for (int j = 0; j < current_NNLayer->num_neurons; j++) {
            double error = 0.0;
            // Sum the weight * delta from the next NNLayer
            for (int k = 0; k < next_NNLayer->num_neurons; k++) {
                error += next_NNLayer->weights[k * next_NNLayer->num_inputs + j] * next_NNLayer->deltas[k];
            }
            current_NNLayer->deltas[j] = error * sigmoid_derivative(current_NNLayer->outputs[j]);
        }
    }

    // 4. Update Weights and Biases
    for (int l = 0; l < net->num_NNLayers; l++) {
        NNLayer *NNLayer = &net->NNLayers[l];
        double *NNLayer_inputs = (l == 0) ? inputs : net->NNLayers[l-1].outputs;

        for (int j = 0; j < NNLayer->num_neurons; j++) {
            NNLayer->biases[j] += learning_rate * NNLayer->deltas[j];
            for (int i = 0; i < NNLayer->num_inputs; i++) {
                NNLayer->weights[j * NNLayer->num_inputs + i] += learning_rate * NNLayer->deltas[j] * NNLayer_inputs[i];
            }
        }
    }
}

// --------------------------------------------------------
// 5. File Saving and Loading
// --------------------------------------------------------

void save_network(Network *net, const char *filename) {
    FILE *file = fopen(filename, "wb");
    if (!file) { perror("Failed to open file for saving"); return; }

    fwrite(&net->num_inputs, sizeof(int), 1, file);
    fwrite(&net->num_NNLayers, sizeof(int), 1, file);

    for (int l = 0; l < net->num_NNLayers; l++) {
        NNLayer *NNLayer = &net->NNLayers[l];
        fwrite(&NNLayer->num_inputs, sizeof(int), 1, file);
        fwrite(&NNLayer->num_neurons, sizeof(int), 1, file);
        fwrite(NNLayer->weights, sizeof(double), NNLayer->num_inputs * NNLayer->num_neurons, file);
        fwrite(NNLayer->biases, sizeof(double), NNLayer->num_neurons, file);
    }

    fclose(file);
    printf("Network saved to %s\n", filename);
}

Network* load_network(const char *filename) {
    FILE *file = fopen(filename, "rb");
    if (!file) { perror("Failed to open file for loading"); return NULL; }

    Network *net = (Network*)malloc(sizeof(Network));
    fread(&net->num_inputs, sizeof(int), 1, file);
    fread(&net->num_NNLayers, sizeof(int), 1, file);

    net->NNLayers = (NNLayer*)malloc(net->num_NNLayers * sizeof(NNLayer));

    for (int l = 0; l < net->num_NNLayers; l++) {
        NNLayer *NNLayer = &net->NNLayers[l];
        fread(&NNLayer->num_inputs, sizeof(int), 1, file);
        fread(&NNLayer->num_neurons, sizeof(int), 1, file);

        // Allocate memory for weights, biases, outputs, and deltas
        NNLayer->weights = (double*)malloc(NNLayer->num_inputs * NNLayer->num_neurons * sizeof(double));
        NNLayer->biases = (double*)malloc(NNLayer->num_neurons * sizeof(double));
        NNLayer->outputs = (double*)malloc(NNLayer->num_neurons * sizeof(double));
        NNLayer->deltas = (double*)malloc(NNLayer->num_neurons * sizeof(double));

        fread(NNLayer->weights, sizeof(double), NNLayer->num_inputs * NNLayer->num_neurons, file);
        fread(NNLayer->biases, sizeof(double), NNLayer->num_neurons, file);

        // Initialize runtime arrays to zero
        for (int i = 0; i < NNLayer->num_neurons; i++) {
            NNLayer->outputs[i] = 0.0;
            NNLayer->deltas[i] = 0.0;
        }
    }

    fclose(file);
    printf("Network loaded from %s\n", filename);
    return net;
}

// --------------------------------------------------------
// 6. Example Usage (Airplane Steering Context)
// --------------------------------------------------------

int CreateNet() {
    srand(time(NULL));

    // Define Network: 3 Inputs, 2 Hidden NNLayers (8 neurons each), 2 Outputs
    // Inputs: [Pitch angle, Roll angle, Altitude error]
    // Outputs: [Elevator command, Aileron command]
    int inputs = 3;
    int hidden_NNLayers = 2;
    int hidden_size = 8;
    int outputs = 2;

    printf("Creating network...\n");
    Network *autopilot = create_network(inputs, hidden_NNLayers, hidden_size, outputs);

    // Dummy training data:
    // If nose is down (negative pitch), pull up (high elevator)
    // Values normalized between 0.0 and 1.0 for the Sigmoid function
    double train_in[3]  = {0.2, 0.5, 0.5}; // Pitch down, wings level, altitude good
    double train_out[2] = {0.8, 0.5};      // Elevator UP, Ailerons neutral

    printf("Training over 10,000 epochs...\n");
    for (int epoch = 0; epoch < 10000; epoch++) {
        backpropagate(autopilot, train_in, train_out, 0.1); // Learning rate = 0.1
    }

    // Test the network
    double *result = feed_forward(autopilot, train_in);
    printf("After training - Target: [%.2f, %.2f] | Output: [%.2f, %.2f]\n",
           train_out[0], train_out[1], result[0], result[1]);

    // Save, Free, and Reload to demonstrate persistence
    save_network(autopilot, "autopilot_model.bin");
    free_network(autopilot);

    Network *loaded_autopilot = load_network("autopilot_model.bin");

    // Test loaded network to verify state
    result = feed_forward(loaded_autopilot, train_in);
    printf("Loaded network Output: [%.2f, %.2f]\n", result[0], result[1]);

    free_network(loaded_autopilot);
    return 0;
}