#include "CbmRichRingSelectAnn.h"
#include "CbmRichRingLight.h"
#include "CbmRichRingSelectImpl.h"

#include "TTree.h"
#include "TSystem.h"
#include "TMultiLayerPerceptron.h"

#include <iostream>

using std::cout;
using std::endl;

CbmRichRingSelectAnn::CbmRichRingSelectAnn ():
   fAnnWeights(""),
   fNN(NULL),
   fSelectImpl(NULL)
{
   fAnnWeights = string(gSystem->Getenv("VMCWORKDIR"))  +"/parameters/rich/rich_select_ann_weights.txt";
}

CbmRichRingSelectAnn::~CbmRichRingSelectAnn()
{

}

void CbmRichRingSelectAnn::Init ()
{
    fSelectImpl = new CbmRichRingSelectImpl();

    TTree *simu = new TTree ("MonteCarlo","MontecarloData");
    Double_t x[6];
    Double_t xOut;

    simu->Branch("x0", &x[0],"x0/D");
    simu->Branch("x1", &x[1],"x1/D");
    simu->Branch("x2", &x[2],"x2/D");
    simu->Branch("x3", &x[3],"x3/D");
    simu->Branch("x4", &x[4],"x4/D");
    simu->Branch("x5", &x[5],"x5/D");
    simu->Branch("xOut", &xOut,"xOut/D");

    fNN = new TMultiLayerPerceptron("x0,x1,x2,x3,x4,x5:10:xOut", simu);
    cout << "-I- CbmRichRingSelectNeuralNet: get NeuralNet weight parameters from: " << fAnnWeights << endl;
    fNN->LoadWeights(fAnnWeights.c_str());

}

void CbmRichRingSelectAnn::DoSelect(
      CbmRichRingLight* ring)
{
    if (ring->GetRadius() >= 10.f || ring->GetRadius() <= 0.f ||
        ring->GetNofHits() <= 5.f ||
        ring->GetRadialPosition() <= 0.f || ring->GetRadialPosition() >= 999.f ){

        ring->SetSelectionNN(-1.f);
        return;
    }
    ring->SetNofHitsOnRing(fSelectImpl->GetNofHitsOnRingCircle(ring));
    if (ring->GetNofHitsOnRing() < 5){
    	ring->SetSelectionNN(-1.f);
    	return;
    }

    ring->SetAngle(fSelectImpl->GetAngle(ring));
    if (ring->GetAngle() < 0.f || ring->GetAngle() > 6.3f){
    	ring->SetSelectionNN(-1.f);
    	return;
    }

    double params[6];
    params[0] = ring->GetNofHits() / 45.;
    params[1] = ring->GetAngle() / 6.28;
    params[2] = ring->GetNofHitsOnRing() / 45.;
    params[3] = ring->GetRadialPosition() / 110.;
    params[4] = ring->GetRadius() / 10.;
    params[5] = (ring->GetChi2() / ring->GetNofHits()) / 0.4;

    float nnEval = fNN->Evaluate(0, params);

    ring->SetSelectionNN(nnEval);
}