SciPost logo

SciPost Submission Page

Multi-scale Mining of Kinematic Distributions with Wavelets

by Ben G. Lillard, Tilman Plehn, Alexis Romero, Tim M. P. Tait

This is not the latest submitted version.

This Submission thread is now published as

Submission summary

Authors (as registered SciPost users): Benjamin Lillard · Tilman Plehn · Tim Tait
Submission information
Preprint Link: https://arxiv.org/abs/1906.10890v2  (pdf)
Date submitted: 2019-08-28 02:00
Submitted by: Lillard, Benjamin
Submitted to: SciPost Physics
Ontological classification
Academic field: Physics
Specialties:
  • High-Energy Physics - Phenomenology
Approaches: Experimental, Theoretical

Abstract

Typical LHC analyses search for local features in kinematic distributions. Assumptions about anomalous patterns limit them to a relatively narrow subset of possible signals. Wavelets extract information from an entire distribution and decompose it at all scales, simultaneously searching for features over a wide range of scales. We propose a systematic wavelet analysis and show how bumps, bump-dip combinations, and oscillatory patterns are extracted. Our kinematic wavelet analysis kit KWAK provides a publicly available framework to analyze and visualize general distributions.

Current status:
Has been resubmitted

Reports on this Submission

Anonymous Report 1 on 2019-9-5 (Invited Report)

  • Cite as: Anonymous, Report on arXiv:1906.10890v2, delivered 2019-09-05, doi: 10.21468/SciPost.Report.1152

Strengths

1- interesting and useful idea
2- wavelets seem more naturally and easily suited to this use than to more complicated uses in the existing literature (jet substructure, pile up, etc)
3- great to make these tools available more widely!

Weaknesses

1- exposition is not very clear (lots of definitions in a somewhat rapid-fire manner, and then some sloppiness when "plugging in" certain definitions)
2- some choices seem to be arbitrary

Report

In general, this is a good idea applied to what seems like a natural use case. I think there are a few specific choices to expand on and justify and a few small corrections, but over all I think it's a very good introductory / exploratory study.

Requested changes

1- the authors should show the signal that has been injected in the top panels of Fig 1 again in the second panel (with proper normalization) so that we can see how well their "10% leading" prescription does -- I think this is pretty important, since otherwise we don't know if this technique actually does recover the signal. Perhaps this is a functionality that needs to be built into KWAK. (Incidentally, I also don't see \tilde f_0, which they say is shown.)
2- I'm confused by Eq 6 (which I think is of central importance -- so confusion here is not good!). What is the object f that has no tilde but has a (single) subscript? How is this related to the (well defined) object \tilde f_{l,m} or to f?
3- I confirm Eq 9, but again I'm confused by what \tilde f means here and how it relates to these f's that have single subscripts. I think the subscripts here are different than the ones in point 2, but this is very confusing...
4- "valorous" or "quixotic" are better than "desperate" in the beginning of Sec 3 :)
5- They should explain or justify their "color coding" in Fig 4 and after -- "how significant" is a coefficient to be shown in red?
6- Another key point: how are 3%, 5%, 12% chosen in Fig 4? How sensitive are their results to these choices? And again in Fig 4, it would be great to the see the injected signal as well. (Minor point about Fig 4, they say it extends to 2.6 TeV but appears only to go to ~ 1.5 TeV)
7- When they discuss Kaluza-Klein models, can they comment on other newer models with many particles? e.g. 1902.05535, or their current ref 6
8- The penultimate sentence of Sec 3.2 appears to be missing a word or words
9- In Eq 15, is the lower case p supposed to be P? If not, how are they related?
10- The second part of their Appendix A about the FRGS should be promoted to a section or at least a subsection of Sec 2. This is very interesting and (seemingly) allows for a statistical test of their wavelet analysis without any of the arbitrary numbers like 3%, 5% etc above. Removing such model-dependence is very useful with a new analysis procedure like the one they propose

  • validity: high
  • significance: high
  • originality: high
  • clarity: good
  • formatting: good
  • grammar: perfect

Login to report or comment