MadGraph5_aMC@NLO

How to reduce fraction of events with wgts > 1

Asked by Elzbieta

Hello,

  I am generating processes with 2->4 matrix elements in pp collision:
      gu > gu ta+ ta-,
      gd > gd ta+ ta-
and I have realised that the fraction of events with wgts > 1 as reported in files eg.
  /SubProcesses/P0_gq_gqtaptam/G10/run_01_log.txt

is not negligible, on the level f 2% - 3% . It does not depend on the number of generated events,
i.e. is not decreasing with increasing statistics. It is also random, i.e. redoing generation several
gives each time different fraction (are you using random generator which is machine clock seeded?)

What are the consequences for the precision/stability of generated samples?
Could I configure better initialisation step to reduce it to below 0.1% level?

thanks

Elzbieta

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
Tim Stelzer Edit question
Last query:
Last reply:
Revision history for this message
Tim Stelzer (tstelzer) said : 		#1

There is a parameter in the file run_config.inc trunc_max with default value 0.01, which means the program tries to truncate < 1% of the cross section. Since you are seeing 2-3%, this suggests that during the run, the system is getting 1 or 2 events with very large weights (2% of cross section).

This can be an indication that it is having difficulty with the integration. You may consider increasing the values of min_events and max_events in the run_config file to force the program to use more events in sampling the phase space.

Can you help with this problem?

Provide an answer of your own, or ask Elzbieta for more information if necessary.

To post a message you must log in.