TMinuit operation in the "for" loop

ROOT
1

I am performing fittings using TMinuit in various settings.

Each fitting has different conditions, do I need to define a new TMinuit object? The current code is as follows

double global_factor;

void FCN(int &npar, double *gin, double &f, , double *par, int iflag)
{
     // some process
     f =  par[0] + ... .... .... + global_factor;
}

int main() 
{
    // JUST DEFINE ONE TMinuit, in the begging 
    TMinuit *min = new TMinuit(1);
    min->SetFCN(FCN);
    min->mnparm(0, "local_factor", 0.5, 0.01, 0, 1, ierflg);

    for (int i=0; i<1000; i++) {

         global_factor = // set different global factor

         // Do I need to define a new TMinuit object here???
         // TMinuit *min = new TMinuit(1);
         // min->SetFCN(FCN);
         // min->mnparm(0, "local factor", 0.5, 0.01, 0, 1, ierflg);

         min->mnexcm("MIGRAD", arglist ,2,ierflg);
    }
}

Do you know if the above code would reflect the result of one previous fitting and not correctly implement the fitting of the new loop settings?

Or, even if the previous result is reflected, is it possible to find the minimum parameter for each new setting?

2

Hi,
You don’t need to recreate the TMinuit class for every iteration. You might want to call TMinuit::mncler() for every iteration to reset it.

Lorenzo

4

thanks for the comment!

the main2 below might not be able to find proper minimm?? (because the previous parameter and the fitting results are remain).

int main1()
    // JUST DEFINE ONE TMinuit, in the begging 
    TMinuit *min = new TMinuit(1);
    min->SetFCN(FCN);
    
    for (int i=0; i<1000; i++) {
        global_factor = // set different global factor
        min->mncler(); // clear the parameters
        min->mnparm(0, "local_factor", 0.5, 0.01, 0, 1, ierflg);
        min->mnexcm("MIGRAD", arglist ,2,ierflg);
    }

int main2()
    // JUST DEFINE ONE TMinuit, in the begging 
    TMinuit *min = new TMinuit(1);
    min->SetFCN(FCN);
    
    for (int i=0; i<1000; i++) {
        global_factor = // set different global factor
        //min->mncler(); // clear the parameters
        min->mnparm(0, "local_factor", 0.5, 0.01, 0, 1, ierflg);
        min->mnexcm("MIGRAD", arglist ,2,ierflg);
    }
5

Hello,
I am not sure I have understood your problem. It is possible that without calling mncler works anyway or works even better, since it is re-using some information from the previous minimization.

Lorenzo