[gate-users] uncorrelated random number seeds

Hiroshi Watabe watabe at ri.ncvc.go.jp
Mon Oct 17 11:53:28 CEST 2005


Dear Dirk,

Thank you again for your explanation of the random seeds.
I will definitely try simulation with your seeds.

Hiroshi
ps. StartSeq1.rndm and StartSeq2.rndm  in seeds.zip are identical.

From: Dirk Kruecker <d.kruecker at fz-juelich.de>
Subject: Re: [gate-users] uncorrelated random number seeds
Date: Mon, 17 Oct 2005 11:36:36 +0200
Message-ID: <200510171136.36930.d.kruecker at fz-juelich.de>

> > Dear Dirk,
> >
> > Thank you for your quick response.
> >
> > Although my generated seed files contain totally different numbers (only
> > first line of all files is same as 9876), it is not guarantee for the
> 
> 9876 identifys the stream. If you do a short simulation and save the random 
> status afterwards you will be at a different position in the same stream, say 
> a 1000 or so random numbers later. You may imagine the random number stream 
> as a fixed list of random numbers. Starting at different positions will not 
> give you independent numbers. On the other hand since each simulated event 
> will consume a different amount of random numbers it is not so obvious that 
> such a procedure creates corelated results. Anyhow with the random seeds 
> files I  send you, you on the safe side each one starts an independent stream 
> i.e. a different list of random numbers.
> 
> Hope it helps 
> Dirk 
> 
> > independencies of these files, isn't it?
> >
> > I will try simulations with your seeds.
> >
> > regards,
> >
> > Hiroshi
> >
> >  From: Dirk Kruecker <d.kruecker at fz-juelich.de>
> > Subject: Re: [gate-users] uncorrelated random number seeds
> > Date: Mon, 17 Oct 2005 10:36:46 +0200
> > Message-ID: <200510171036.46083.d.kruecker at fz-juelich.de>
> >
> > > Dear Hiroshi
> > >
> > > the method  Mutanga Theodore described does not create independent random
> > > seeds but in most cases this should not be a real problem.  Maybe there
> > > is some other reason for your results but it is difficult to say what the
> > > reason is for the correlations you see. I posted a detailed explaination
> > > how to handle the random files on this list at Jun 30, 2005. I also
> > > include a set of random seed files that are indepent for sure.
> > >
> > > Best regards
> > > Ditrk
> > >
> > > Am Montag, 17. Oktober 2005 09:01 schrieb Hiroshi Watabe:
> > > > Dear Gate-users,
> > > >
> > > > I try to do 4 simulations with different seeds. In the past gate-users
> > > > talk, I found the article (posted by Mutanga Theodore on 17 Dec 2004)
> > > > which explain how to generate different seeds files.
> > > > As the article explains, I did mock simulation with 10 Bq source and
> > > > generated four different endOfRun.rndm. By using the command
> > > > /random/resetEngineFrom, I initialized random seeds.
> > > > I supposed to get four different results but four simulations produced
> > > > very similar results. I say "similar" because I got exactly same
> > > > projection data but the root outputs were slightly different in each
> > > > others. In the root file, I found only event ID and time
> > > > were different but else (for instance, source position and global
> > > > position) were exactly same in Singles trees (I'm doing SPECT
> > > > simulation).
> > > >
> > > > Any suggestions and advice are appreciate.
> > > >
> > > > Hiroshi
> > > >
> > > > --------------------------------------------------
> > > >              Hiroshi Watabe, PhD
> > > >
> > > > watabe at ri.ncvc.go.jp
> > > > Department of Investigative Radiology
> > > > Advanced Medical Engineering Center
> > > > National Cardiovascular Center Research Institute
> > > >
> > > > 5-7-1 Fujishirodai, Suita, Osaka 565-8565
> > > >
> > > > Tel:06-6833-5012-2560 or 06-6833-5004-2560
> > > > Fax:06-6835-5429
> > > > --------------------------------------------------
> > > >
> > > > _______________________________________________
> > > > gate-users mailing list
> > > > gate-users at lphe1pet1.epfl.ch
> > > > http://lphe1pet1.epfl.ch/mailman/listinfo/gate-users
> >
> > _______________________________________________
> > gate-users mailing list
> > gate-users at lphe1pet1.epfl.ch
> > http://lphe1pet1.epfl.ch/mailman/listinfo/gate-users
> 
> _______________________________________________
> gate-users mailing list
> gate-users at lphe1pet1.epfl.ch
> http://lphe1pet1.epfl.ch/mailman/listinfo/gate-users


More information about the Gate-users mailing list