ROOT::Math::Minimizer quastion

[mtn63]

Dear Root experts!

I have two questions aboutROOT::Math::Minimizer class:

minimization provides a vector of parameters x [0], x [1] …
in my case the SetLimitedVariable [1] depends on x [0], that is to say : at each iteration I have a new x [0], my limits ss_low, ss_up (see the example below /* … */) depends on the value x [0] given by the minimization.

Can I do a second “min->Minimize();” after recovering my parameters, something like what I write below

ROOT::Math::Minimizer* min = ROOT::Math::Factory::CreateMinimizer("Minuit2", "Migrad");
 min->SetMaxFunctionCalls(1000000);//1000000
    min->SetMaxIterations(10000);
    min->SetTolerance(0.001);
    min->SetPrintLevel(3);//3 au lieu de -1 pour afficher tout
   
    ROOT::Math::Functor f(&Fcn, 4);
.....
min->SetLimitedVariable(0,"ll",  variable[0], step[0], ll_low, ll_up);
min->SetLimitedVariable(1,"ss",  variable[1], step[1],ss_low, ss_up);
....
 // do the minimization   
    min->Minimize();
    const double *xs = min->X();

cout<<"par[0] ="<< xs[0]<<" !par[1] ="<< xs[1]....

/* I want to do somthing like: 
ss_low = xs[0] - 3;
ss_up = xs[0] + 3;
min->SetLimitedVariable(0,"ll",  variable[0], step[0], ll_low, ll_up);
min->SetLimitedVariable(1,"ss",  variable[1], step[1],ss_low, ss_up);
min->Minimize(); */

Thank you very much in advance,

[moneta]

Hi,

It should work as you have written. If you have any problems, please let me know

Cheers

Lorenzo

[mtn63]

@mtn63

// do the minimization min->Minimize(); const double xs = min->X(); cout<<“par[0] =”<< xs[0]<<" !par[1] ="<< xs[1]… / I want to do somthing like: ss_low = xs[0] - 3; ss_up = xs[0] + 3; min->SetLimitedVariable(0,“ll”, variable[0], step[0], ll_low, ll_up); min->SetLimitedVariable(1,“ss”, variable[1], step[1],ss_low, ss_up); min->Minimize(); */

Thank you for your answer Lorenzo,
so there is no problem if I make two minimizations as below? (First method)
what I want to do is get the variable [0] after the first iteration and use it in the variable [1] like what I write in the second method (see below), is that possible?:

//first method

--------------------------------------------------------------------

//first minimization
    min->Minimize();
    const double *xs = min->X();

cout<<"par[0] ="<< xs[0]<<" !par[1] ="<< xs[1]....


//retrieve parameters   
ss_low = xs[0] - 3;
ss_up = xs[0] + 3;
min->SetLimitedVariable(0,"ll",  variable[0], step[0], ll_low, ll_up);
min->SetLimitedVariable(1,"ss",  variable[1], step[1],ss_low, ss_up);


//second minimization
min->Minimize(); 

---

///second method

ROOT::Math::Minimizer* min = ROOT::Math::Factory::CreateMinimizer("Minuit2", "Migrad");
 min->SetMaxFunctionCalls(1000000);//1000000
    min->SetMaxIterations(10000);
    min->SetTolerance(0.001);
    min->SetPrintLevel(3);//3 au lieu de -1 pour afficher tout
   
    ROOT::Math::Functor f(&Fcn, 4);
.....
min->SetLimitedVariable(0,"ll",  variable[0], step[0], ll_low, ll_up);
ss_low=min->X()[0] - 3;
ss_up=min->X()[0] + 3;
min->SetLimitedVariable(1,"ss",  variable[1], step[1],ss_low, ss_up);
....

 // do the minimization   
    min->Minimize();
    const double *xs = min->X();

cout<<"par[0] ="<< xs[0]<<" !par[1] ="<< xs[1]....

[mtn63]

Hi Lorenzo,
I obtain this result and the Minimization did NOT converge !!! , I have 6 free parameters,
have you any explanation please?

Minuit2Minimizer: Minimize with max-calls 1000000 convergence for edm < 0.001 strategy 1
MnSeedGenerator: for initial parameters FCN = 0.2042598768742
MnSeedGenerator: Initial state:   - FCN =  0.2042598768742 Edm = 0.0529160919757376 NCalls =     25
MnSeedGenerator: Negative G2 found - new state:   - FCN =  0.1194703058954 Edm = 8.08352977319165e-08 NCalls =    197
VariableMetric: start iterating until Edm is < 2e-06
VariableMetric: Initial state   - FCN =  0.1194703058954 Edm = 8.08352977319165e-08 NCalls =    197
VariableMetric: Iteration #   0 - FCN =  0.1194703058954 Edm = 8.08352977319165e-08 NCalls =    197
VariableMetric: Iteration #   1 - FCN =  0.1194701824964 Edm = 5.13330807132362e-08 NCalls =    224
Info in <Minuit2>: MnHesse: 2nd derivative zero for Parameter  : name = dist_12
Info in <Minuit2>: MnHesse fails and will return diagonal matrix 
VariableMetric: After Hessian   - FCN =  0.1194701824964 Edm = 5.13330807132362e-08 NCalls =    229
VariableMetric: Iteration #   2 - FCN =  0.1194701824964 Edm = 5.13330807132362e-08 NCalls =    229
Number of iterations 3
----------> Iteration 0
           FVAL = 0.119470305895 Edm = 8.08352977319e-08 Nfcn = 197
           Error matrix change = 1
           Parameters :  p0 = 13 p1 = 7.50000014787254 p2 = 8.73013145502175 p3 = 5.33246923288788 p4 = 15 p5 = 4
----------> Iteration 1
           FVAL = 0.119470182496 Edm = 5.13330807132e-08 Nfcn = 224
           Error matrix change = 0.503243359804529
           Parameters :  p0 = 13 p1 = 7.50000005056474 p2 = 8.73012951736379 p3 = 5.33247330876207 p4 = 15.000000601616 p5 = 4.00006974190362
----------> Iteration 2
           FVAL = 0.119470182496 Edm = 5.13330807132e-08 Nfcn = 229
           Error matrix change = 1
           Parameters :  p0 = 13 p1 = 7.50000005056474 p2 = 8.73012951736379 p3 = 5.33247330876207 p4 = 15.000000601616 p5 = 4.00006974190362
Info in <Minuit2>: Minuit2Minimizer::Minimize : Minimization did NOT converge, Hesse is not valid
Minuit2Minimizer : Invalid Minimum - status = 2
FVAL  = 0.119470182496389
Edm   = 5.13330807132362e-08
Nfcn  = 229