Нейронка не обучается

void Net::backward(elementSet* _idealSet)
{
    //Recount output delta
    for (int i = 0; i < layers[layersCount - 1].count; ++i)
    {
        layers[layersCount - 1].neuron[i].recountOutputDelta(_idealSet->at(i));
    }
    //Recount hidden delta
    for (int layNum = layersCount - 2; layNum > 0; --layNum)
    {
        for (int neuNum = 0; neuNum < layers[layNum].count; ++neuNum)
        {
            layers[layNum].neuron[neuNum].recountHiddenDelta(&layers[layNum+1], neuNum);
        }
    }
 
    //Update Weight
    for (int layNum = layersCount - 1; layNum > 0; --layNum)
    {
        for (int neuNum = 0; neuNum < layers[layNum].count; ++neuNum)
        {
            layers[layNum].neuron[neuNum].updateWeight(trainSpeed, moment);
        }
    }
}
void Neuron::updateWeight(double _speed, double _moment)
{
    std::vector<sinaps>::iterator connect = connections.begin();
    double lastModWeight = 0;
 
    for (; connect < connections.end(); ++connect)
    {
        lastModWeight = _speed * delta * connect->first->getOut() + lastModWeight * _moment;
        connect->second += lastModWeight;
    }
}
double Neuron::recountOutputDelta(double _idealOut)
{
    delta = (_idealOut - out) * derivateActivation();
    return delta;
}
 
double Neuron::recountHiddenDelta(Layer* _nextLayer, int _myNum)
{
    double sum = 0;
    for (int i = 0; i < _nextLayer->count; ++i)
    {
        sum += _nextLayer->neuron[i].delta * _nextLayer->neuron[i].connections[_myNum].second;
    }
    delta = sum * derivateActivation();
    return delta;
}

Layer* forward(elementSet* _set)
    {
        if (_set->size() != *(topology)) throw "Invalid element's set! Set's size != size of topology's input layer.\n";
 
        //Set input 
        std::vector<double>::const_iterator elem = _set->cbegin();
        for (int j = 0; j < *(topology); ++j)
        {
            layers[0].neuron[j].setInput( &(*elem) );
            ++elem;
        }
 
        //Forward
        for (int i = 1; i < layersCount; ++i)
        {
            for (int j = 0; j < *(topology + i); ++j)
            {
                layers[i].neuron[j].activation();
            }
        }
 
        return &layers[layersCount - 1];
    }