Minuit2 NaNs after awhile

Hello all,

        So I'm attempting to do a log likelihood minimization and everything goes well for awhile and then NaN values get introduced suddenly and everything goes wrong. I investigated a little and things go to NaN values when they have no way to do so. No dividing my zero or anything. I do not know why this happens. I've read that this could be a step size issue? I'm using ROOT 5.34, OpenMP and I'm on OSX Mavericks as well if you think those are important. Thank you very much

Hi,

NaN are normally created by the function provided by the user to be minimised. Are you sure you are not for example computing the log of a negative number ?
I would suggest you to print in your function for every time is called the input parameter values and the computed function value. You would see when the first NaN appears and for which reason. You could also try to use an higher print level (e.g. 3) in Minuit2

Best Regards

Lorenzo