Re: [ROOT] mnsave and TMinuit

From: Rene Brun (Rene.Brun@cern.ch)
Date: Fri Jan 30 2004 - 22:37:37 MET


John,

Minuit mnsave was essentially saving the current parameters and 
covariance matrix. You can achieve the same result by writing
the TMinuit object to a file and restore it if necessary.
To save the TMinuit object (eg pointed by gMinuit) do
  TFile f("MinuitSave.root","recreate");
  gMinuit->Write();

To reat it back
  TFile f("MinuitSave.root")
  TMinuit *minuit = (TMinuit*)f.Get("MINUIT");
  minuit->Dump(); //to see the current members

I have just a stupid question of principle.
After 4 weeks, do you still believe in the numerical stability of the 
results ?

Rene Brun

On Fri, 30 
Jan 2004, Dr. John Krane wrote:

> 
> Hi everyone,
> 
> I'm running jobs with TMinuit that take 1-4 weeks to complete.  I am finding that system instabilities can cause my machine's uptime to be less than the time required to finish the fit.  It is *very frustrating* to wait a long time and end up with nothing.
> 
> I see that TMinuitOld purportedly has a method called mnsave that writes the intermediate fit results to file "in format which can be reread by Minuit for restarting".  That sounds like exactly what I want to happen every few hours, but I cannot find the file TMinuitOld to write a version that does so.
> 
> Has anybody on this list put together a version of TMinuit that writes a journal-file and can restart a fit from said file?  If not, does anybody have any advice for me as I try to write my own?  I don't at this time know what variables TMinuit would need to save and recover, or if any other classes would also have to recover parameter values.
> 
>     - John
> 
> ---
> Dr. John Krane
> 
> 
> ________________________________________________________________
> The best thing to hit the Internet in years - NetZero HiSpeed!
> Surf the Web up to FIVE TIMES FASTER!
> Only $14.95/ month -visit www.netzero.com to sign up today!
> 



This archive was generated by hypermail 2b29 : Sun Jan 02 2005 - 05:50:05 MET