-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.c
455 lines (383 loc) · 14.8 KB
/
main.c
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
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
// Deep Neural Network serial code with 2 layers
// Names: Anuradha Agarwal, Thomas Keller, Zack Humphries
// Final Project
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include "evaluation.h"
int numInputs;
int numHiddenNodes;
int numHiddenNodes2;
double learningRate;
int numberOfEpochs;
#define numOutputs 1 // Number of outputs
#define numTrain 455 // Number of rows in train set
#define numTest 114 // Number of rows in test set
#define numTrainingSets 569 // Number of instances of total data
// The main starts here
int main(int argc, char *argv[]) {
// Checking for command line arguments
if (argc != 4){
printf("Please provide number of inputs as your first argument, learning rate as your second argument and number of epochs as your third argument. \n");
exit(1);
}
numInputs = atoi(argv[1]); // Number of columns
numHiddenNodes = atoi(argv[1]); // Number of nodes in the first hidden layer
numHiddenNodes2 = atoi(argv[1]); // Number of nodes in the second hidden layer
learningRate = atof(argv[2]); // Constant for learning rate
numberOfEpochs = atoi(argv[3]); // Number of epochs
char training_set[100];
char testing_set[100];
int characters;
// Updating variables depending on the command line arguments
if (numInputs == 30){
strncpy(training_set, "datasets/train_data.csv", sizeof(training_set));
strncpy(testing_set, "datasets/test_data.csv", sizeof(testing_set));
characters = 1024;
}
else if (numInputs == 4800){
strncpy(training_set, "datasets/train_data4801.csv", sizeof(training_set));
strncpy(testing_set, "datasets/test_data4801.csv", sizeof(testing_set));
characters = 1602400;
}
else{
printf("Please make sure your first argument is either 30 or 4800. \n");
exit(2);
}
// Learning rate
const double lr = learningRate;
// Declare pointers for dynamically allocated arrays
double* hiddenLayer;
double* hiddenLayer2;
double* outputLayer;
double* hiddenLayerBias;
double* hiddenLayerBias2;
double* outputLayerBias;
double** hiddenWeights;
double** hiddenWeights2;
double** outputWeights;
// Allocate memory for hidden layer nodes vectors
hiddenLayer = (double*) malloc(numHiddenNodes * sizeof(double));
hiddenLayer2 = (double*) malloc(numHiddenNodes2 * sizeof(double));
outputLayer = (double*) malloc(numOutputs * sizeof(double));
// Allocate memory for hidden layer bias vectors
hiddenLayerBias = (double*) malloc(numHiddenNodes * sizeof(double));
hiddenLayerBias2 = (double*) malloc(numHiddenNodes2 * sizeof(double));
outputLayerBias = (double*) malloc(numOutputs * sizeof(double));
// Allocate memory for hidden and output weights matrices
hiddenWeights = (double**) malloc(numInputs * sizeof(double*));
hiddenWeights2 = (double**) malloc(numInputs * sizeof(double*));
outputWeights = (double**) malloc(numHiddenNodes2 * sizeof(double*));
// Dynamically allocate memory for hiddenWeights and hiddenWeights2
for (int i = 0; i < numInputs; i++) {
hiddenWeights[i] = (double*) malloc(numHiddenNodes * sizeof(double));
hiddenWeights2[i] = (double*) malloc(numHiddenNodes2 * sizeof(double));
}
// Dynamically allocate memory for outputWeights
for (int i = 0; i < numHiddenNodes2; i++) {
outputWeights[i] = (double*) malloc(numOutputs * sizeof(double));
}
// Dynamically allocate memory for training and testing inputs and outputs
double **trainingInputs = (double **)malloc(numTrain * sizeof(double *));
for (int i = 0; i < numTrain; i++) {
trainingInputs[i] = (double *)malloc(numInputs * sizeof(double));
}
// Dynamically allocate memory for testingInputs
double **testingInputs = (double **)malloc(numTest * sizeof(double *));
for (int i = 0; i < numTest; i++) {
testingInputs[i] = (double *)malloc(numInputs * sizeof(double));
}
// Dynamically allocate memory for trainingInputs
double **trainingOutputs = (double **)malloc(numTrain * sizeof(double *));
for (int i = 0; i < numTrain; i++) {
trainingOutputs[i] = (double *)malloc(numOutputs * sizeof(double));
}
// Allocate memory for testingOutputs
double *testingOutputs = (double *)malloc(numTest * sizeof(double));
// Reading csv files
char buffer[characters];
char buffer2[characters];
char *record, *line;
char *record2, *line2;
int i = 0, j = 0;
double inputTrain[numTrain][numInputs+1];
double inputTest[numTrain][numInputs+1];
// Read Train data from train_data.csv
FILE *fstream = fopen(training_set, "r");
if (fstream == NULL) {
printf("\n file opening failed train ");
return -1;
}
while ((line = fgets(buffer, sizeof(buffer), fstream)) != NULL) {
record = strtok(line, ",");
while (record != NULL) {
inputTrain[i][j++] = strtod(record, NULL);
record = strtok(NULL, ",");
}
if (j == numInputs)
i += 1;
}
fclose(fstream);
i = 0, j = 0;
// Read Test data from test_data.csv
FILE *gstream = fopen(testing_set, "r");
if (gstream == NULL) {
printf("\n file opening failed test ");
return -1;
}
while ((line2 = fgets(buffer2, sizeof(buffer2), gstream)) != NULL) {
record2 = strtok(line2, ",");
while (record2 != NULL) {
inputTest[i][j++] = strtod(record2, NULL);
record2 = strtok(NULL, ",");
}
if (j == numInputs)
i += 1;
}
fclose(gstream);
// Training data (inputs)
for (int ro=0; ro<numTrain; ro++)
{
for(int columns=1; columns<numInputs+1; columns++)
{
trainingInputs[ro][columns-1] = inputTrain[ro][columns];
}
}
// Testing data (inputs)
for (int ro=0; ro<numTest; ro++)
{
for(int columns=1; columns<numInputs+1; columns++)
{
testingInputs[ro][columns-1] = inputTest[ro][columns];
}
}
// Training data (outputs)
for (int ro=0; ro<numTrain; ro++)
{
for(int columns=0; columns<1; columns++)
{
trainingOutputs[ro][columns] = inputTrain[ro][columns];
}
}
// Testing data (outputs)
for (int ro=0; ro<numTest; ro++)
{
for(int columns=0; columns<1; columns++)
{
testingOutputs[ro] = inputTest[ro][columns];
}
}
// Initialize bias and weight terms to random
// Hidden layer 1 weights
for(int i = 0; i < numHiddenNodes; i++){
hiddenLayerBias[i] = initWeights();
}
// Hidden layer 2 weights
for(int i = 0; i < numHiddenNodes2; i++){
hiddenLayerBias2[i] = initWeights();
}
// Output layer weights
for(int i = 0; i < numOutputs; i++){
outputLayerBias[i] = initWeights();
}
// Hidden layer 1 bias
for(int i = 0; i < numInputs; i++){
for(int j = 0; j < numHiddenNodes; j++){
hiddenWeights[i][j] = initWeights();
}
}
// Hidden layer 2 bias
for(int i = 0; i < numHiddenNodes; i++){
for(int j = 0; j < numHiddenNodes2; j++){
hiddenWeights2[i][j] = initWeights();
}
}
// Output layer bias
for(int i = 0 ; i < numHiddenNodes; i++){
for(int j = 0; j < numOutputs; j++){
outputWeights[i][j] = initWeights();
}
}
// Specify training set
int trainingSetOrder[numTrain];
for(int i = 0 ; i < numTrain ; i++)
{
trainingSetOrder[i] = i;
}
// Start time measurement
clock_t start, end;
start = clock();
// Training loop
for(int epoch = 0; epoch < numberOfEpochs; epoch++){
shuffle(trainingSetOrder, numTrain);
for(int x = 0; x < numTrain; x ++){
int i = trainingSetOrder[x];
// forward pass: compute hidden layer activation
// hidden layer 1
for(int j =0; j < numHiddenNodes; j++){
double activation = hiddenLayerBias[j];
for(int k = 0; k < numInputs; k++){
activation += trainingInputs[i][k] * hiddenWeights[k][j];
}
hiddenLayer[j] = relu(activation);
}
// hidden layer 2
for(int j =0; j < numHiddenNodes2; j++){
double activation = hiddenLayerBias2[j];
for(int k = 0; k < numHiddenNodes; k++){
activation += hiddenLayer[k] * hiddenWeights2[k][j];
}
hiddenLayer2[j] = relu(activation);
}
// compute output layer activation
for(int j =0; j < numOutputs; j++){
double activation = outputLayerBias[j];
for(int k = 0; k < numHiddenNodes2; k++){
activation += hiddenLayer2[k] * outputWeights[k][j];
}
outputLayer[j] = sigmoid(activation);
}
// Print training output (only first 6 inputs for readability)
printf("Input: %g | %g | %g | %g | %g | %g | Output: %g Expected Output: %g \n",
trainingInputs[i][1], trainingInputs[i][2], trainingInputs[i][3], trainingInputs[i][4], trainingInputs[i][5], trainingInputs[i][6],
outputLayer[0], trainingOutputs[i][0]);
// Backpropagation
// Compute change in output weights
double deltaOutput[numOutputs];
for(int j = 0; j < numOutputs; j++){
double error = (trainingOutputs[i][j] - outputLayer[j]); // L1
deltaOutput[j] = error * dSigmoid(outputLayer[j]) ;
}
// Compute change in hidden weights (second layer)
double deltaHidden2[numHiddenNodes2];
for(int j = 0; j < numHiddenNodes2; j++){
double error = 0.0f;
for(int k = 0; k < numOutputs; k++){
error += deltaOutput[k] * outputWeights[j][k];
}
deltaHidden2[j] = error * dRelu(hiddenLayer[j]);
}
// Compute change in hidden weights (first layer)
double deltaHidden[numHiddenNodes];
for(int j = 0; j < numHiddenNodes; j++){
double error = 0.0f;
for(int k = 0; k < numHiddenNodes2; k++){
error += deltaHidden2[k] * hiddenWeights2[j][k];
}
deltaHidden[j] = error * dRelu(hiddenLayer2[j]);
}
// Apply change in output weights
for(int j = 0; j < numOutputs; j++){
outputLayerBias[j] += deltaOutput[j] * lr;
for(int k = 0; k < numHiddenNodes2; k++){
outputWeights[k][j] += hiddenLayer2[k] * deltaOutput[j] * lr;
}
}
// Apply change in second hidden layer weights
for(int j = 0; j < numHiddenNodes2; j++){
hiddenLayerBias[j] += deltaHidden[j] * lr;
for(int k = 0; k < numHiddenNodes; k++){
hiddenWeights2[k][j] += hiddenLayer[k] * deltaHidden2[j] * lr;
}
}
// Apply change in first hidden layer weights
for(int j = 0; j < numHiddenNodes; j++){
hiddenLayerBias[j] += deltaHidden[j] * lr;
for(int k = 0; k < numInputs; k++){
hiddenWeights[k][j] += trainingInputs[i][k] * deltaHidden[j] * lr;
}
}
}
}
end = clock();
/*
// print final weights after done training
fputs ("\nFinal Hidden Weights\n[ ", stdout);
for(int j = 0; j < numHiddenNodes; j++){
fputs ("[ ", stdout);
for(int k = 0; k < numInputs; k++){
printf("%f ", hiddenWeights[k][j]);
}
fputs("] ", stdout);
}
fputs ("\nFinal Hidden2 Weights\n[ ", stdout);
for(int j = 0; j < numHiddenNodes2; j++){
fputs ("[ ", stdout);
for(int k = 0; k < numHiddenNodes; k++){
printf("%f ", hiddenWeights2[k][j]);
}
fputs("] ", stdout);
}
fputs ("]\nFinal Hidden Biases\n[ ", stdout);
for(int j = 0; j < numHiddenNodes; j++){
printf("%f ", hiddenLayerBias[j]);
}
fputs ("]\nFinal Hidden2 Biases\n[ ", stdout);
for(int j = 0; j < numHiddenNodes2; j++){
printf("%f ", hiddenLayerBias2[j]);
}
fputs ("]\nFinal Output Weights\n", stdout);
for(int j = 0; j < numOutputs; j++){
fputs ("[ ", stdout);
for(int k = 0; k < numHiddenNodes2; k++){
printf("%f ", outputWeights[k][j]);
}
fputs("] \n", stdout);
}
fputs ("\nFinal Output Biases\n[ ", stdout);
for(int j = 0; j < numOutputs; j++){
printf("%f ", outputLayerBias[j]);
}
fputs("] \n", stdout);
*/
// Building neural network with the trained weights and bias
// initialize testInput and testResults
double testInput[numTest];
double testResults[numTest];
// looping through the matrix and sending in one vector at a time to evaluate
for(int i = 0; i < numTest; i++) {
for (int j = 0; j < numInputs; j++) {
testInput[j] = testingInputs[i][j];
}
// predicted solution
testResults[i] = evaluation(numInputs, numHiddenNodes, numHiddenNodes2, numOutputs,
testInput,hiddenWeights,hiddenWeights2,outputWeights,hiddenLayerBias,hiddenLayerBias2,outputLayerBias);
printf("predicted results: %f actual result: %f \n", testResults[i], testingOutputs[i]);
}
accuracy(testResults,testingOutputs,numTest); // accuracy, precision, fscore
// calculate total time in ms
double duration = ((double)end - start)/CLOCKS_PER_SEC;
printf("Total time: %fs \n", duration); // time
for (int i = 0; i < numInputs; i++) {
free(hiddenWeights[i]);
free(hiddenWeights2[i]);
}
for (int i = 0; i < numHiddenNodes2; i++) {
free(outputWeights[i]);
}
free(hiddenLayer);
free(hiddenLayer2);
free(outputLayer);
free(hiddenLayerBias);
free(hiddenLayerBias2);
free(outputLayerBias);
free(hiddenWeights);
free(hiddenWeights2);
free(outputWeights);
// Free dynamically allocated memory for training and testing inputs and outputs
for (int i = 0; i < numTrain; i++) {
free(trainingInputs[i]);
}
free(trainingInputs);
for (int i = 0; i < numTest; i++) {
free(testingInputs[i]);
}
free(testingInputs);
for (int i = 0; i < numTrain; i++) {
free(trainingOutputs[i]);
}
free(trainingOutputs);
free(testingOutputs);
return 0;
}