User Tools

Site Tools


2023:groups:smhiggs:higgs-cell-resample:start

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
2023:groups:smhiggs:higgs-cell-resample:start [2023/06/16 11:46]
stephen.jones
2023:groups:smhiggs:higgs-cell-resample:start [2023/06/18 16:23] (current)
stephen.jones
Line 1: Line 1:
-===== Negative Weight Mitigation for Higgs Signal/​Background =====+===== Negative Weight Mitigation ​(also for Higgs Signal/​Background?) ===== 
 + 
 +**Code**: [[https://​github.com/​a-maier/​cres|cres]] 
 + 
 +== Specific Topics == 
 +  * Apply cell resampling ([[https://​arxiv.org/​abs/​2109.07851|paper 1]] [[https://​arxiv.org/​abs/​2303.15246|paper 2]] [[https://​github.com/​a-maier/​cres|source]] [[https://​cres.hepforge.org/​|binaries]]) to Higgs signal/​background (closure tests, measure of simplification e.g. #event reduction) 
 +  * Graph of fraction of negative weights for Higgs processes/​background (slide 5/9 [[https://​phystev.cnrs.fr/​wiki/​_media/​2023:​cell_resampling.pdf|Jeppe talk]]) 
 + 
 +== Processes == 
 +  * di-photon NLO sample (m_gamgam centred around Higgs mass window)
  
 == General Topics == == General Topics ==
-  * Graph of fraction of negative weights for Higgs processes/​background (slide 5/9 Jeppe talk) 
   * Discussion on good metrics (what to compare before considering events close)   * Discussion on good metrics (what to compare before considering events close)
   * Measure the impact on reals vs virtuals, which events are most likely to be altered and by how much?   * Measure the impact on reals vs virtuals, which events are most likely to be altered and by how much?
Line 11: Line 19:
   * Can narrow weight distribution for more efficient event unweighting   * Can narrow weight distribution for more efficient event unweighting
  
-== Discussion (winner: most questions per talk) == +== Talk Discussion (winner: most questions per talk) == 
-  * q: why is w+5 improving more than z+3? +  * Q: why is w+5 improving more than z+3?  
-  - initial event sample size? (no, both have 1e9) +    - initial event sample size? (no, both have 1e9) 
-  - dipole cut vs improvement?​ (check dipole cut used) +    - dipole cut vs improvement?​ (check dipole cut used) 
- +  * Q: calculating density of events to "​pre-check"​ that this algorithm may work? 
-- q: calculating density of events to "​pre-check"​ that this algorithm may work? +  * Q: metric needs to be compatible with how ps generator is populating ps, proof that this is not biasing anything? 
- +  * Q: should there be a bias towards positive values (since you iteratively add nearest event until wi>​=0)?​ 
-- q: metric needs to be compatible with how ps generator is populating ps, proof that this is not biasing anything? +    ​- ​ps generator dependent because this alters which events are clustered, can this have an impact on the physics? 
-- q: should there be a bias towards positive values (since you iteratively add nearest event until wi>​=0)?​ +  * Q: plotting statistical uncertainties on the original sample  
-  ​* ​ps generator dependent because this alters which events are clustered, can this have an impact on the physics? +    ​- ​more easily allows verifying that differences are within statistical uncertainty 
- +  * Q: on which type of event does this have the biggest impact? 
-- q: plotting statistical uncertainties on the original sample +    ​- ​imagine reals are more impacted than virtual  
-  ​* ​more easily allows verifying that differences are within statistical uncertainty +    ​- ​we were completely agnostic 
- +  * Q: can we really prove that this does not alter distributions?​ (prove that you preserve distributional structure of the observables you compute) 
-- q: on which type of event does this have the biggest impact? +    ​- ​we do not alter any of the event kinematics 
-  ​* ​imagine reals are more impacted than virtual +    ​- ​there is no cross talk of events separated by more than the maximum allowed distance 
-  ​* ​we were completely agnostic +    ​- ​IRC safe measure important, but this is not a sufficient condition, can you reproduce infrared sensitive observables (e.g. Sudakov shoulder, 0-bin of ptZ), can your smearing reproduce this feature in the limit that the smearing goes to 0? 
- +  * Q: mean/​median/​width of the cell resampling bins 
-- q: can we really prove that this does not alter distributions?​ (prove that you preserve distributional structure of the observables you compute) +    ​- ​we have plots that we can examine
-  ​* ​we do not alter any of the event kinematics +
-  ​* ​there is no cross talk of events separated by more than the maximum allowed distance +
-  ​* ​IRC safe measure important, but this is not a sufficient condition, can you reproduce infrared sensitive observables (e.g. Sudakov shoulder, 0-bin of ptZ), can your smearing reproduce this feature in the limit that the smearing goes to 0? +
- +
-- q: mean/​median/​width of the cell resampling bins +
-  ​* ​we have plots which we can examine +
- +
- +
-== Specific Topics == +
-  * Apply cell resampling ([[https://​arxiv.org/​abs/​2109.07851|paper 1]] [[https://​arxiv.org/​abs/​2303.15246|paper 2]] [[https://​github.com/​a-maier/​cres|source]] [[https://​cres.hepforge.org/​|binaries]]) to Higgs signal/​background (closure tests, measure of simplification e.g. #event reduction) +
- +
-== Processes == +
-  * di-photon NLO sample (m_gamgam centred around Higgs mass window) +
  
2023/groups/smhiggs/higgs-cell-resample/start.1686908766.txt.gz · Last modified: 2023/06/16 11:46 by stephen.jones