[RMCProfile-users] Automated weight optimization function

Tomas Northam tom.northam at gmail.com
Thu Jan 12 08:04:13 GMT 2023

 Dear Viktor,

Thank you for the response.

I think I understand most of it. The only thing that I do not understand
(and maybe it is something very obvious) is the difference between the
total residual and the chi square. Obviously, as the fit to the data get
better, both decrease. But if the residual is calculated differently than
chi, does the chi square minimization have preference over the total
residual? So could there be a case where a set of moves decreases chi
square and not the total residual (and would thus be a problem)?

Kind regards,

On Wed, 11 Jan 2023 at 15:58, Krayzman, Victor L. (Assoc) via
rmcprofile-users <rmcprofile-users at shadow.nd.rl.ac.uk> wrote:

> RMCProfile uses numerous data sets and restraints of quite different
> types. Respectively, the residuals for each type of data set and restraint
> are calculated in different ways. The procedure of computations of each
> particular residual is important but has no direct relation to the
> automated weight optimization. Residuals are positive numbers which go to
> zero when the calculated data approaches to the experimental one.
> You are right in general that RMC minimizes the total residual but to be
> more precise it employs the Metropolis algorithm which accepts all
> favorable  moves and also may accept some moves which increase the total
> residual. The probability of acceptance of those moves depends of the value
> of total residual gain after a move. So, the global weight rescaling
> parameter T allows to vary this probability and therefore keep the total
> percentage of accepted move at desired level.
>  Victor Krayzman
> -----Original Message-----
> From: rmcprofile-users <rmcprofile-users-bounces at shadow.nd.rl.ac.uk> On
> Behalf Of Tomas Northam via rmcprofile-users
> Sent: Wednesday, January 11, 2023 5:00 AM
> To: rmcprofile-users at rmcprofile.org
> Cc: Tomas Northam <tom.northam at gmail.com>
> Subject: [RMCProfile-users] Automated weight optimization function
> Dear all,
> I have a question regarding the *Automated weight optimization* function.
> Two questions:
> 1. How does it calculate the residual for each data set? And how is the
> global weight rescaling parameter T chosen?
> 2. How does the total residual come into the fitting? I understand that
> RMC works on a chi minimization procedure, so Ido not understand how the *
> Automated weight optimization* would affect the final fitting.
> Thank you for the help.
> Kind regards,
> Tom.
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <
> https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.rmcprofile.org%2Fpipermail%2Frmcprofile-users%2Fattachments%2F20230111%2F6b5c8cfe%2Fattachment.htm&data=05%7C01%7Cvictor.krayzman%40nist.gov%7Ca0ceff1142ed47ca13af08daf3bd1225%7C2ab5d82fd8fa4797a93e054655c61dec%7C1%7C0%7C638090290636694676%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=SuugWg16ZLdwuJ00tgJIfQL%2BAag7XstH3ieKl3trrx4%3D&reserved=0
> >
> _______________________________________________
> rmcprofile-users mailing list
> rmcprofile-users at rmcprofile.org
> https://gcc02.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.rmcprofile.org%2Fmailman%2Flistinfo%2Frmcprofile-users&data=05%7C01%7Cvictor.krayzman%40nist.gov%7Ca0ceff1142ed47ca13af08daf3bd1225%7C2ab5d82fd8fa4797a93e054655c61dec%7C1%7C0%7C638090290636694676%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=4oUpAKIcg%2FO%2B2uGto9wHwiS0q0k6UZt43e98qv3vi6Y%3D&reserved=0
> _______________________________________________
> rmcprofile-users mailing list
> rmcprofile-users at rmcprofile.org
> http://lists.rmcprofile.org/mailman/listinfo/rmcprofile-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.rmcprofile.org/pipermail/rmcprofile-users/attachments/20230112/1628c309/attachment.htm>

More information about the rmcprofile-users mailing list