SciPost Submission Page
Picking the low-hanging fruit: testing new physics at scale with active learning
by Juan Rocamonde, Louie Corpe, Gustavs Zilgalvis, Maria Avramidou, Jon Butterworth
This Submission thread is now published as
|Authors (as registered SciPost users):
|Jonathan Butterworth · Louie Corpe · Juan Rocamonde
|SciPost Physics Core
Since the discovery of the Higgs boson, testing the many possible extensions to the Standard Model has become a key challenge in particle physics. This paper discusses a new method for predicting the compatibility of new physics theories with existing experimental data from particle colliders. By using machine learning, the technique obtained comparable results to previous methods (>90% precision and recall) with only a fraction of their computing resources (<10%). This makes it possible to test models that were impossible to probe before, and allows for large-scale testing of new physics theories.
Published as SciPost Phys. 13, 002 (2022)
Submission & Refereeing History
You are currently on this page
Reports on this Submission
- Cite as: Anonymous, Report on arXiv:2202.05882v1, delivered 2022-03-09, doi: 10.21468/SciPost.Report.4649
4) Reasonably comprehensive for this kind of thing
6) Substantive progress in the field
This is an excellent paper that meets the journal's criteria. I recommend it be published. There is very little to criticize because it is lucid and complete.
I only spotted a few typos:
Page 3, second-to-last paragraph:
-- "and the results distribution are compared to the observed results for LHC analyses"
Page 4, first full paragraph"
Page 8, last paragraph:
Page 9, second paragraph:
Page 11, last partial paragraph:
-- "this make sense"
- Cite as: Anonymous, Report on arXiv:2202.05882v1, delivered 2022-03-08, doi: 10.21468/SciPost.Report.4640
1- To my knowledge, this paper is tackling an issue which is definitely overlooked in our field. It is something that deserves more attention.
2- The paper is very well written and easy to read.
3- Adding a novel model for testing was a good idea
No major weaknesses.
1- The paper lacks (in some parts more than in others) a comprehensive set of citations. For authors expert in the field, as these are, it's not much work since you do not have to find the sources, you have them and you usually fix this problem very easily by spending a boring 2 hours filling in all the places where a citation is missing.
2- The paper was clearly proof-read but it requires another two or three runs. There are still grammar errors and typos and some figures are missing a proper caption
This is a very interesting paper on an interesting and relevant subject written by someone that knows the English language. Well done. Being a HEP physicist myself I find the paper well written and easy to understand. It requires some editorial work detailed below but I would say nothing major.
The main part that I would add is a proper and detailed study of the uncertainties related to the contours that you are proposing e.g. in figure 9. E.g. in GAMBIT it is well knows that they use a Poissonian likelihood marginalised over a rescaling parameter to account for systematic uncertainties. What do you do?
Another point that I would like to raise is your dependence on the RIVET project. If I understand correctly, you need an analysis to be on RIVET in order to use you tool on it. This is a major shift in the tone you are using in Section 1. While I understand the importance of getting the reader to understand that you code "matters" I think the relatively small ratio of ATLAS, CMS, and LHCb analyses available on RIVET should be acknowledged as a bottleneck.
For the rest, well done and thanks!
The changes that are outlined in the weakness and report sections:
1- Can you try to add a paragraph about uncertainty handlings or the lack of it? E.g. like GAMBIT use of a Poissonian likelihood marginalised over a rescaling parameter to account for systematic uncertainties. If you do not have this I am not saying you should modify your code but acknowledge the issue.
1- Increase citations (very obvious in Sec 1)
2- Tone done a bit Sec 1 or explain the problem with RIVET availability
3- Proofread another two or three times, there are typos and grammatical errors, you are clearly native speakers so this should not take long
4- I would remove Figure 1 which is not adding any information (but do as you like)
5- You need better Figure captioning across the paper
6- Avoid footnotes
7- Equations should be numbered
8- In some figure’s axes are not labelled
9- Section 3.4 is a bit too qualitative to my taste, here a good quantitative paragraph about systematic uncertainties could be a welcomed addition