precision control and parameter scan in batch mode

Asked by Yong Du on 2021-03-24

Hi,

Using CalcHEP in the batch mode, I am wondering if there are any ways to control the precision of the cross section. Say, I want CalcHEP to stop the next round iteration when the cross section precision is better than 1%, is there a way to achieve this in the batch mode? In principle I can set a very large number of simulation to improve the precision, but it would be a waste of the computing sources if the calculation turns our to converge efficiently.

My second question is related to the parameter scan. I can do this in the batch mode without any problem. But it looks to me that the scan range of a certain parameter, say the center-of-mass energy, can only change linearly since the step size can only be set to a constant. I am wondering if there is a way to do it logarithmically since the scan range in my case changes by orders of magnitude (from O(1) to O(1e5)).

Thank you in advance,
Yong

Question information

Language:
English Edit question
Status:
Solved
For:
CalcHEP Edit question
Assignee:
No assignee Edit question
Solved by:
Alexander Belyaev
Solved:
Last query:
Last reply:
Revision history for this message
Alexander Belyaev (alexander.belyaev) said :
#1

There are situations (whihc happen quite often) when the accuracy 1% can not be reached because of the settings user make, in this case if there would be possibility to set up 1% accuracy, the program will be caught into the infinite loop --- and this is the least what you want.

In case of simple 2->2 or 2->3 processes 10K/100K will be enough to reach 1%
in most of the cases. For 2->5,2->6 you would likely need 1M calls,
The number of call you can easily control in batch or gui, I guess you know how to do this.
If not, pelase let us know.

There is no universal solution/recipe to the problem you have raised.
If the type of the process you study is the same for the loop then you can find very quickly how many calls do you need to optimise it and have 1% preicsion.

If you need more specific advice for the process you are deadling with -- just specify your process.

Revision history for this message
Yong Du (fallive) said :
#2

Dear Alexander,

Thank you a lot! I got your point and I do know how to control that in the batch mode. Thank you!

Now, for my second question (see below), do you have any suggestion?

"My second question is related to the parameter scan. I can do this in the batch mode without any problem. But it looks to me that the scan range of a certain parameter, say the center-of-mass energy, can only change linearly since the step size can only be set to a constant. I am wondering if there is a way to do it logarithmically since the scan range in my case changes by orders of magnitude (from O(1) to O(1e5))."

Best,
Yong

Revision history for this message
Best Alexander Belyaev (alexander.belyaev) said :
#4

Sorry, I have realised that you have asked two questions here
from the beginning.
Here is the answer on your second question.
For non-linear scan you can use the following trick.

You can make scan over any variable(s) which does not belong to thre model
and construct the function of it.
For example

1) For Model parmeter:

Scan parameter : var
Scan begin : 1
Scan step size : 1
Scan n steps : 3
Parameter: Mh=10**var

In this case Mh will take vslues 10,100,1000

2) For collider energy

p1: 10**var/2
p2: 10**var/2

Scan parameter : var
Scan begin : 3
Scan step size : 1
Scan n steps : 3

In this case Mh will take values sqrts will take eventually 1TeV, 10TeV abd 100 TeV vlues

Using this method you can construct virtually any scan you like ---
one- and multi-dimensional one

Revision history for this message
Yong Du (fallive) said :
#5

Dear Alexander,

Thank you very much for the suggestion!

Regarding my second question, I actually read the manual carefully about performing such a parameter scan, but failed to find any information about how to introduce some user-defined/external functions for that purpose. My method for doing this was quite nasty, and previously I thought I need to use bash script to interfere with "calchep_batch" such that a logarithmic scan would be possible. Now I am glad to know that it turns out to be much simpler to do this as you wrote above.

Therefore, for the benefit of future users, I would suggest including what you wrote here (maybe also include some info about how to apply some user-defined functions for a custom scan?) in the manual of CalcHEP, I think someone like me would really appreciate it.

Thank you again and best wishes,
Yong

Revision history for this message
Yong Du (fallive) said :
#6

Thanks Alexander Belyaev, that solved my question.

Revision history for this message
Alexander Belyaev (alexander.belyaev) said :
#7

Dear Yong,

thank you. You are right.
This description will be included in the new CalcHEP
paper which is coming soon.

Best wishes,
Alexander

On 25/03/2021 10:01, Yong Du wrote:
> Question #696232 on CalcHEP changed:
> https://answers.launchpad.net/calchep/+question/696232
>
> Yong Du posted a new comment:
> Dear Alexander,
>
> Thank you very much for the suggestion!
>
> Regarding my second question, I actually read the manual carefully about
> performing such a parameter scan, but failed to find any information
> about how to introduce some user-defined/external functions for that
> purpose. My method for doing this was quite nasty, and previously I
> thought I need to use bash script to interfere with "calchep_batch" such
> that a logarithmic scan would be possible. Now I am glad to know that it
> turns out to be much simpler to do this as you wrote above.
>
> Therefore, for the benefit of future users, I would suggest including
> what you wrote here (maybe also include some info about how to apply
> some user-defined functions for a custom scan?) in the manual of
> CalcHEP, I think someone like me would really appreciate it.
>
> Thank you again and best wishes,
> Yong
>

--
______________________________________________________________________
Prof. Alexander S Belyaev (<email address hidden>)
https://www.hep.phys.soton.ac.uk/content/alexander-belyaev

School of Physics & Astronomy, University of Southampton
Office 5047, SO17 1BJ, TEL: +44 23805 98509, FAX: +44 23805 93910
.....................................................................
Particle Physics Department, Rutherford Appleton Laboratory
Didcot, OX11 0QX, TEL: +44 12354 45562, FAX: +44 12354 46733
.....................................................................
CERN, CH-1211 Geneva 23, Switzerland
Office 40/1-B20, Mailbox: E27910, TEL: +41 2276 71642
______________________________________________________________________