void FramesWeightFrame::mousePressEvent(QMouseEvent* mouseEvent)
{
mouseDown = true;
weightedFrame = mouseEvent->x()/FrameBarWidth;
adjustWeight(mouseEvent->y());
repaint();
}
void FramesWeightFrame::mouseMoveEvent(QMouseEvent* mouseEvent)
{
if(mouseDown)
{
weightedFrame = mouseEvent->x()/FrameBarWidth;
if(weightedFrame<0 || weightedFrame>=frameCount())
return;
adjustWeight(mouseEvent->y());
repaint();
}
}
void BpAlgSt::start(){
Q_ASSERT(net != NULL);
Q_ASSERT(data != NULL);
emit started();
//value initialization
running = true;
actIter = 1;
actTime = 0;
timer.restart();
double sumErr = 0;
for(int i = 0; i < data->minPatternCount(); i++){
output = net->layerOutput(data->inputVector(i));
sumErr += calcError(i);
}
actError = sumErr;
emit update(0, actTime, actError);
//learning main cycle
while(running){
for(int i = 0; i < data->minPatternCount(); i++){
//feedforward
output = net->layerOutput(data->inputVector(i));
//output layer delta calculation
calcOutputDelta(i);
//inner layer delta calculation
calcInnerDelta();
//weight adjustment
adjustWeight();
}
//current time
actTime = timer.elapsed();
//output error calculation
double sumErr = 0;
for(int i = 0; i < data->minPatternCount(); i++){
output = net->layerOutput(data->inputVector(i));
sumErr += calcError(i);
}
actError = sumErr;
//emits update signal once per each update interval
if(actIter % updateInterv == 0){
emit update(actIter, actTime, actError);
}
//stop conditions
if(actTime >= stopTimeVal) break;
if(actIter >= stopIter) break;
if(actError <= stopErrorVal) break;
actIter++;
}
//running flag to false
running = false;
//signal that tells that learning is finished
emit stoped();
}
