void Backpropagation::trainOnlineCV(Mlp& network,
MATRIX& trainingInputs,
VECTOR& trainingTargets,
MATRIX& testInputs,
VECTOR& testTargets)
{
VECTOR trainingOutputs(trainingTargets.size(), 0.0);
VECTOR testOutputs(testTargets.size(), 0.0);
while(error > tolerance && testCount < maxTestCount)
{
VECTOR::iterator output = trainingOutputs.begin();
VECTOR::iterator target = trainingTargets.begin();
for(MATRIX::iterator input = trainingInputs.begin();
input != trainingInputs.end();
++input, ++target, ++output)
{
*output = network(*input);
double err = *output - *target;
getWeightUpdates(network, *input, err);
applyWeightUpdates(network);
++iteration;
if(iteration >= maxIterations)
break;
}
++epoch;
error = mse(trainingTargets, trainingOutputs);
// Early-stopping using test (cross-validation) error
testOutputs = network(testInputs);
testError = mse(testTargets, testOutputs);
if(testError < minTestError)
{
// Preserve test error and network weights
minTestError = testError;
W = network.W;
V = network.V;
biasW = network.biasW;
biasV = network.biasV;
testCount = 0;
}
else
{
++testCount;
}
}
network.W = W;
network.V = V;
network.biasW = biasW;
network.biasV = biasV;
testError = minTestError;
}
int main() {
ImportarDados();
InicializarVariaveisGradiente();
//NormalizarDados();
printf("\nCusto inicial igual a : %f", mse());
GradienteDescendente();
printf("\nCusto final igual a : %f", mse());
printf("\nTheta zero: %f e theta 1: %f",theta[0],theta[1]);
return 0;
}
void YuvAnalyse::psnr(YuvFrame& a, YuvFrame& b, float* psnrY, float* psnrU, float* psnrV){
float mseY, mseU, mseV;
mse(a, b, &mseY, &mseU, &mseV);
*psnrY = 10.0 * log10f((255.0*255.0)/mseY);
*psnrU = 10.0 * log10f((255.0*255.0)/mseU);
*psnrV = 10.0 * log10f((255.0*255.0)/mseV);
}
//detecting mouse button
void dtcbttn ()
{
/* function 1 returns 0 until a key is pressed */
//while (_bios_keybrd(_KEYBRD_READY) == 0) //while (!kbhit () )
{
int x,y;
static int mov_flag=0;
float x1,y1;
static int xx,yy;
static buffer[2];
in.x.ax = 3;
int86 (0X33,&in,&out);
//clrscr();
//gotoxy(40,30);
x = out.x.cx;
y = out.x.dx;
//printf ("\nLeft || X - %d Y - %d", x, y);
x1=x/639.0*80;
y1=y/399.0*50;
if (x1>(int)x1) x=(int)x1+1;
else x=x1;
if (y1>(int)y1) y=(int)y1+1;
else y=y1;
/*if (mov_flag==1 && ((xx!=x) ||(yy!=y)))
{puttext(xx,yy,xx,yy,buffer); printf("%d",buffer[0]);}
mov_flag=1;
xx=x;
yy=y;
gettext(xx,yy,xx,yy,buffer);
*/
if (out.x.bx == 1)
{
//gotoxy(2,2);
//cprintf ("\nLeft || X - %f Y - %f",x1,y1 );
//cprintf ("\nLeft || X - %d Y - %d", x, y);
//gotoxy(x,y);
//cprintf("||");
mse_flag=1;
mse(x,y);
}
else if (out.x.bx == 2)
{
//printf ("\nRight");
delay (100);}
else if (out.x.bx == 3)
{//printf ("\nMiddle");
delay (100);}
else
mse_flag=0;
// Otherwise due to quick computer response 100s of words will get print
}
delay(100);
}
Void snr (const CVideoObjectPlane& vop1, const CVideoObjectPlane& vop2, Double dsnr [3])
{
Double msError [3];
mse (vop1, vop2, msError);
for (UInt i = 0; i < 3; i++) {
if (msError [i] == 0.0)
dsnr [i] = 1000.0;
else
dsnr [i] = (log10 (255 * 255 / msError [i]) * 10.0);
}
}
// Try to train the net to learn the given function. It returns the number
// of adjustments needed to learn the function, or FAILURE if it was never
// learned. The variable "stats" tells whether or not to print information
// while training.
int PDP::trial(Function* target, int stat, double rate)
{
if(target->get_num_inputs() != num_inputs) {
cerr << "number of inputs for network and function do not match" << endl;
exit(1);
}
int freq = 10; // How often do we want results printed?
double weights[MAX_WEIGHTS];
for(int x = 0; x < num_weights; x++)
weights[x] = (rand() % 10000) / 5000.0; // From 0 to 2
for(int n = 0; n < MAX_ADJUSTMENTS && !success(target, weights); n++) {
if(!stat && (n % freq == 0))
cout << "Adjustments: " << n
<< " Error: " << mse(target, weights) << endl;
gradient(target, weights, rate);
}
if(!stat) {
if(n < MAX_ADJUSTMENTS) {
cout << "Weights:";
for(x = 0; x < num_weights; x++)
cout << " " << weights[x];
cout << endl;
} else
cout << "Function not learned in " << MAX_ADJUSTMENTS
<< " adjustments" << endl;
}
if(n < MAX_ADJUSTMENTS)
return n;
else
return FAILURE;
}
inline double
psnr(unsigned char *p1, unsigned char *p2, size_t length, size_t maxval)
{
return 10*log10(maxval*maxval/mse(p1,p2,length));
}
inline double
rmse(unsigned char *p1, unsigned char *p2, size_t length)
{
return sqrt(mse(p1,p2,length));
}
