// Инициализация фильтра Кальмана
public KalmanFilter()
{
m_timeLastMeasurement = clock();
int dynam_params = 4; // x,y,dx,dy
int measure_params = 2;
m_pKalmanFilter = cvCreateKalman(dynam_params, measure_params);
state = cvCreateMat( dynam_params, 1, CV_32FC1 );
// Генератор случайных чисел
cvRandInit( &rng, 0, 1, -1, CV_RAND_UNI );
cvRandSetRange( &rng, 0, 1, 0 );
rng.disttype = CV_RAND_NORMAL;
cvRand( &rng, state );
process_noise = cvCreateMat( dynam_params, 1, CV_32FC1 ); // (w_k)
measurement = cvCreateMat( measure_params, 1, CV_32FC1 ); // two parameters for x,y (z_k)
measurement_noise = cvCreateMat( measure_params, 1, CV_32FC1 ); // two parameters for x,y (v_k)
cvZero(measurement);
// F matrix data
// F is transition matrix. It relates how the states interact
// For single input fixed velocity the new value
// depends on the previous value and velocity- hence 1 0 1 0
// on top line. Новое значение скорости не зависит от предыдущего
// значения координаты, и зависит только от предыдущей скорости- поэтому 0 1 0 1 on second row
const float[] F = {
1, 0, 1, 0,//x + dx
0, 1, 0, 1,//y + dy
0, 0, 1, 0,//dx = dx
0, 0, 0, 1,//dy = dy
};
memcpy( m_pKalmanFilter->transition_matrix->data.fl, F, sizeof(F));
cvSetIdentity( m_pKalmanFilter->measurement_matrix, cvRealScalar(1) ); // (H)
cvSetIdentity( m_pKalmanFilter->process_noise_cov, cvRealScalar(1e-5) ); // (Q)
cvSetIdentity( m_pKalmanFilter->measurement_noise_cov, cvRealScalar(1e-1) ); // (R)
cvSetIdentity( m_pKalmanFilter->error_cov_post, cvRealScalar(1));
// choose random initial state
cvRand( &rng, m_pKalmanFilter->state_post );
//InitializeCriticalSection(&mutexPrediction);
}
static void Main(string[] args)
{
int[][] X_train = new int[NUMBER_OF_EXAMPLES][];
int[] y_train = new int[NUMBER_OF_EXAMPLES];
for (var i = 0; i <= X_train.Length; i++)
{
X_train[i] = new int[Constants.FEATURES];
}
int[][] X_test = new int[NUMBER_OF_EXAMPLES][];
int[] y_test = new int[NUMBER_OF_EXAMPLES];
for (var i = 0; i <= X_test.Length; i++)
{
X_test[i] = new int[Constants.FEATURES];
}
ReadFiles();
var mcTsetlinMachine = new MultiClassTsetlinMachine();
float average = 0.0f;
for (var i = 0; i < 1000; i++)
{
mcTsetlinMachine.Initialize();
TimeSpan start_total = 0;
mc_tm_fit(mc_tsetlin_machine, X_train, y_train, NUMBER_OF_EXAMPLES, 200, 3.9);
clock_t end_total = clock();
double time_used = ((double)(end_total - start_total)) / CLOCKS_PER_SEC;
printf("EPOCH %d TIME: %f\n", i + 1, time_used);
average += mc_tm_evaluate(mc_tsetlin_machine, X_test, y_test, NUMBER_OF_EXAMPLES);
printf("Average accuracy: %f\n", average / (i + 1));
}
}
public SidTuneInfoImpl()
{
m_formatString = "N/A";
m_songs = 0;
m_startSong = 0;
m_currentSong = 0;
m_songSpeed = SPEED_VBI;
m_clockSpeed = clock_t.CLOCK_UNKNOWN;
m_compatibility = compatibility_t.COMPATIBILITY_C64;
m_dataFileLen = 0;
m_c64dataLen = 0;
m_loadAddr = 0;
m_initAddr = 0;
m_playAddr = 0;
m_relocStartPage = 0;
m_relocPages = 0;
m_fixLoad = false;
m_sidModels = new List <model_t>();
m_sidModels.Add(model_t.SIDMODEL_UNKNOWN);
m_sidChipAddresses = new List <UInt16>();
m_sidChipAddresses.Add(0xd400);
}
