void test_aritificial_neural_network(void){
u32 i;
double fan_in;
boolean successed = FALSE;
initiate_neural_network();
for (i = 0; i < MAX_LOOP_TIMES; i++){
train_neural_network();
if (sum_of_error_per_round < TARGET_ERROR_VALUE){
successed = TRUE;
break;
}
if ((i & 2047) == 0)
printf("Variance of diff between real and desired : %f\n",
sum_of_error_per_round);
get_delta();
modify_network_weight();
}
if (successed){
while (TRUE){
printf("Input the number to be calculated. ");
printf("0 for terminated, number should between 0 and 1.\n");
scanf("%lf", &fan_in);
if (fan_in){
if ((fan_in > 0) && (fan_in < 1))
printf("Input: %f, desired: %f, real: %f\n",
fan_in, proceed(fan_in), cos(fan_in));
else
printf("Invalid input, should between 0 and 1.\n");
}
else
break;
}
} else {
printf("Error! Difference between real output and desired: %f\n",
sum_of_error_per_round);
}
}
int
main (int argc, char** argv)
{
/*
int option;
while (1)
{
static struct option long_options[] =
{
{"verbose", no_argument, &verbose_flag, 1},
{"brief", no_argument, &verbose_flag, 0},
{"add", no_argument, 0, 'a'},
{"append", no_argument, 0, 'b'},
{"delete", required_argument, 0, 'd'},
{"create", required_argument, 0, 'c'},
{"train", required_argument, 0, 't'},
{0, 0, 0, 0}
};
int option_index = 0;
option = getopt_long (argc, argv, "abc:d:f:",
long_options, &option_index);
if (option == -1)
break;
switch (option)
{
case 0:
if (long_options[option_index].flag != 0)
break;
printf ("option %s", long_options[option_index].name);
if (optarg)
printf (" with arg %s", optarg);
printf ("\n");
break;
case 'a':
puts ("option -a\n");
break;
case 'b':
puts ("option -b\n");
break;
case 'c':
printf ("option -c with value `%s'\n", optarg);
break;
case 'd':
printf ("option -d with value `%s'\n", optarg);
break;
case 'f':
printf ("option -f with value `%s'\n", optarg);
break;
case '?':
break;
default:
abort ();
}
}
if (verbose_flag)
puts ("verbose flag is set");
if (optind < argc)
{
printf ("non-option ARGV-elements: ");
while (optind < argc)
printf ("%s ", argv[optind++]);
putchar ('\n');
}
*/
(void) argc;
(void) argv;
time_series_data_t* tsd = construct_time_series_data("libcsv/test.csv");
struct tm fromt;
struct tm tot;
memset(&fromt, 0, sizeof(struct tm));
memset(&tot, 0, sizeof(struct tm));
fromt.tm_year = 2010 - 1900;
fromt.tm_mon = 1 - 1;
fromt.tm_mday = 1;
tot.tm_year = 2010 - 1900;
tot.tm_mon = 3 - 1;
tot.tm_mday = 1;
generate_training_set_files_from_time_series_data (tsd, mktime(&fromt), mktime(&tot), 5, 2, "snp500.in", "snp500.out");
destruct_time_series_data(tsd);
size_t config[] = {30,35,12};
//size_t config[] = {2,4,1};
neural_network_t* nn = construct_neural_network(config, 3, -2.0, 2.0, &initialize_nguyen_widrow_weights);
training_t* training = construct_training(nn, &elliott_activation, &elliott_derivative, false);
resilient_propagation_data_t* rprop_data = construct_resilient_propagation_data(nn);
// training_set_t* ts = construct_training_set("xor.in", "xor.out");
training_set_t* ts = construct_training_set("snp500.in", "snp500.out");
normalize_training_set(ts);
//debug_training_set(ts);
printf("Final error rate: %g\n", train_neural_network(training, nn, ts, &resilient_propagation_loop, rprop_data, 20000));
save_neural_network_weights(nn, "snp500.weights");
// load_neural_network_weights(nn, "lol");
// printf("Final error rate: %g\n", train_neural_network(training, nn, ts, &resilient_propagation_loop, rprop_data, 20000));
// destruct_neural_network(nn);
// nn = construct_neural_network_from_file("lol");
// printf("Final error rate: %g\n", train_neural_network(training, nn, ts, &resilient_propagation_loop, rprop_data, 20000));
destruct_training(training, nn);
destruct_resilient_propagation_data(rprop_data, nn);
destruct_training_set(ts);
destruct_neural_network(nn);
return 0;
}
