SciPost Submission Page
Semi-visible jets, energy-based models, and self-supervision
by Luigi Favaro, Michael Krämer, Tanmoy Modak, Tilman Plehn, Jan Rüschkamp
This is not the latest submitted version.
Submission summary
| Authors (as registered SciPost users): | Luigi Favaro · Tilman Plehn |
| Submission information | |
|---|---|
| Preprint Link: | scipost_202312_00024v3 (pdf) |
| Code repository: | https://github.com/luigifvr/dark-clr |
| Data repository: | https://zenodo.org/records/12801842 |
| Date submitted: | 2024-11-28 22:33 |
| Submitted by: | Favaro, Luigi |
| Submitted to: | SciPost Physics |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
Abstract
We present DarkCLR, a novel framework for detecting semi-visible jets at the LHC. DarkCLR uses a self-supervised contrastive-learning approach to create observables that are approximately invariant under relevant transformations. We use background-enhanced data to create a sensitive representation and evaluate the representations using a CLR-inspired anomaly score and a normalized autoencoder as density estimators. Our results show a remarkable sensitivity for a wide range of semi-visible jets and are more robust than a supervised classifier trained on a specific signal.
Author indications on fulfilling journal expectations
- Provide a novel and synergetic link between different research areas.
- Open a new pathway in an existing or a new research direction, with clear potential for multi-pronged follow-up work
- Detail a groundbreaking theoretical/experimental/computational discovery
- Present a breakthrough on a previously-identified and long-standing research stumbling block
Author comments upon resubmission
List of changes
1. We use calorimeter towers and tracks as obtained from the eflow algorithm in Delphes. This is now included after the jet clustering paragraph.
2. Fixed.
3. This was already included in the previous comment and is now added to the manuscript.
4. In principle, the Softmax can be applied to multiple dimensions of the input tensor. The index was used to specify which dimension was normalized. Indeed it is a notation abuse. For clarity, we removed the index and specified it in the text.
5. We discuss here the motivations for introducing s_CLR, which are now adopted in the text. Changing the norm of the representations is an effective way of reducing the CLR alignment loss without collapsing the phase space onto a single point. Therefore, we expect information to be encoded in this quantity.
This is an open research topic which can be mathematically proven, see for example [55]. Indeed, this is not enough to ensure that anomalies will have a large norm, and we resolve the ambiguity by applying the L2 regularisation only to the representations of background data.
We did not observe worse convergence after introducing this term. Future studies can try to do the opposite, i.e. reduce the norm of the representation of augmented jets. In this work, we avoided introducing a term which depends on anomalous augmentations.
6. The beginning of the paragraph explains how the p_Z is obtained, and we further discuss the two terms in Eq.(9).
7. Fixed.
8. Changed to d_h for the representation before the head network and d_z for the final unnormalized output vector.
Current status:
Reports on this Submission
Strengths
The paper describes a new tagging algorithm for identifying semi-visible jets
based on a contrastive learning representation.
1 - The algorithm presented is relevant, since it relies on minimal physical features of the signal.
2 - The algorithm performance can be superior to other standard techniques and more model independent
3 - The text is well written and the results support the paper claims.
4 - The paper is clear and accessible to non-experts on current machine learning developments.
Report
The paper clarity has been improved and the issues raised in previous reports have been addressed. Therefore I recommend it for publication.
Recommendation
Publish (surpasses expectations and criteria for this Journal; among top 10%)
Strengths
1- New state-of-the-art technique to search for new signals producing semi-visible jets
2- Method robust against model parameters of the signal
3- Test the sensitivity of the method to hyperparameter choices
4- Method potentially insensitive to simulation choice
Report
The paper introduces an innovative method for tagging semi-visible jets.
The results are shown to be insensitive to variations in BSM parameters compared to supervised or unsupervised learning techniques.
Requested changes
1- In equation (4), "(W^V x)_j" should be written "(W^V x_j)" to be consistent with the rest of the equation
Recommendation
Ask for minor revision