/**
* Learning of all internal receptive field with given
* predictor input vector and scalar output result
*/
inline void learning(const InputVector& input,
scalar output)
{
std::cout << "========== LEARN " << input.transpose() << " " << output << std::endl; //TODO
//Forward learnign to all receptive fields
scalar maxWeight = 0.0;
for (size_t i=0;i<_receptiveFields.size();i++) {
scalar weight = _receptiveFields[i].activationWeight(input);
std::cout << "=== Learn " << i << " with " << weight << std::endl; //TODO
_receptiveFields[i].learning(input, output, weight);
if (weight > maxWeight) {
maxWeight = weight;
}
}
//Create new receptive field if no one have been
//activated more than a threshold
if (maxWeight < _weightGen) {
_receptiveFields.push_back(ReceptiveField<inputDim, scalar>(
_forgettingRate, _penaltyShape,
_gradientLearningRate, input, _shapeInit));
}
}
