Hybrid Quantum Vision Transformers for Event Classification in High Energy Physics
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsFirst of all, I think the manuscript is well written and the study is well done, congratulations to the authors, here are some comments from my side:
L72, ref of 22 should be in L63,
L82: the arrive time is not used, then it is not necessary to show the plots at Fig. 1 (bottom).
L71, please clarify if there is or how to you deal with the photon conversion event in this dataset?
L128, what is the purpose of this normalization? If I understand correctly, there are 2(3) components, image of deposit energies, position embedding, (class token), none of them need normalization.
Fig. 5, why you have another Rx compared to the Ref. 24, and why the total number of parameters is 3dh+1?
Fig. 7,8, why the hybrid class token model performs so bad ? how you implement the class token for hybrid model?
Table 1, please harmonize the positional encoding and positional embedding, and it is not strange to have such conclusion at L215, the position information should already be used in the linear embedding in Fig. 2, so what is the point to do such test, i.e. w/ and w/o positional embedding? The numbers at last row last volume should be “4785/4585”
Conclusion: can you comment on the training or inference time for these models?
Author Response
Please see the attached file.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis is a very well-written and clear article introducing a hybrid vision transformer model to identify photons and electrons within a high energy electromagnetic shower. The hybrid model is applied to simulated data from the CERN CMS Experiment, and the results are compared with a classical vision transformer model. We have only a few suggestions to improve the clarity of the work, mainly to help the general reader.
Line 57
The paper —> This paper
we present —> we introduce
Line 60
we show our —> we present our
and discuss them —> and discuss the implications
we discuss future —> we consider future
Lines 64-65
The CMS detector also records heavy-ion data (p-Pb and Pb-Pb collisions)
https://twiki.cern.ch/twiki/bin/view/CMSPublic/PhysicsResultsHIN
and pp data at lower collision energy:
https://cds.cern.ch/record/2235781/files/LUM-16-001-pas.pdf
Please correct the wording of the sentence to reflect that among the available CMS datasets, you have chosen to study pp data at 13.6TeV
Line 69
Consider including the name(s) and reference(s) for the CMS particle tracker(s)
The term “tracker” is somewhat casual to a reader that might not be in the high energy physics field...maybe just refer the reader to this CMS detector reference: https://inspirehep.net/literature/796887
Line 75
“orthogonal to the beam line” —> in the xy-plane, see Figure X
Consider including the following diagram of the CMS coordinate system as Figure X:
https://wiki.physik.uzh.ch/cms/_detail/latex:cms_coordinate_system.png?id=latex%3Aexample_spherical_coordinates
Line 76
Please define the pseudorapidity eta using the polar angle theta shown in Figure X
Please define the azimuthal angle as phi in Figure X
Figure 1 caption
“The titles list the true labels (real electron or real photon), as well as the corresponding labels predicted by one of the benchmark classical models.” —> “The titles list the true labels (real electron or real photon), as well as the corresponding labels predicted by one of the benchmark classical models (see text for more details).”
Please define “true label”, “real electron”, and “real photon” in the text
Line 92
Please add a citation to a general description/overview of an electromagnetic shower
“causing the e- shower profiles” —> “causing the electromagnetic shower profiles (citation)” For example, this CMS thesis here: https://inspirehep.net/files/1473595af7405b5757d8a2e610c7246d
Line 182
“In addition, MLP layer inside each encoder layer“ —> “Similar to the classical model, the MLP layer inside each hybrid encoder layer” or something similar
Line 200
The “Training Process” section is only one paragraph in length... It might be helpful to expand on this process in the “Discussion” section. For example, a discussion of what the authors tried that did not work as well, and why?
Author Response
Please see the attached file.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsDear authors,
thanks for the reply and the revised manuscript, I think it is in good shape and recommend to publish it.