SciPost Submission Page
Applications of Machine Learning in Constraining Multi-Scalar Models
by Darius Jurčiukonis
Submission summary
| Authors (as registered SciPost users): | Darius Jurčiukonis |
| Submission information | |
|---|---|
| Preprint Link: | https://arxiv.org/abs/2509.24092v1 (pdf) |
| Code repository: | https://github.com/jurciukonis/ML-for-multiples |
| Date submitted: | Oct. 1, 2025, 10:14 a.m. |
| Submitted by: | Darius Jurčiukonis |
| Submitted to: | SciPost Physics Proceedings |
| Proceedings issue: | The 2nd European AI for Fundamental Physics Conference (EuCAIFCon2025) |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approach: | Phenomenological |
Abstract
Machine learning techniques are used to predict theoretical constraints such as unitarity and boundedness from below in extensions of the Standard Model. This approach has proven effective for models incorporating additional SU(2) scalar multiplets, in particular the quadruplet and sixplet cases. High predictive performance is achieved through the use of suitable neural network architectures and well-prepared training datasets. Moreover, machine learning provides a substantial computational advantage, enabling significantly faster evaluations compared to scalar potential minimization.
Current status:
Reports on this Submission
Report #1 by Anonymous (Referee 1) on 2025-12-10 (Invited Report)
The referee discloses that the following generative AI tools have been used in the preparation of this report:
By means of generative artifices (ChatGPT and Gemini), the reviewer’s rough notes were rendered into prose of more orderly disposition; yet the discernment, judgment, and scholarly responsibility rest wholly with the reviewer, upon whose mind alone these deliberations have been wrought.
Strengths
- Clear and motivated application of ML to a physics problem where substantial speedup is achievable.
- The availability of open-source code is very valuable for reproducibility.
Weaknesses
- The input to the neural networks is not described explicitly.
- No comment on the false positive rate, although this is relevant for a filtering-based workflow.
- Some aspects of the presentation could be improved for clarity (e.g., hardware list, Table 1).
Report
Requested changes
- Please specify which quantities form the 10-dimensional input to the networks.
- Add a short comment on the false positive rate to contextualize the filtering performance.
- Replace or shorten the bullet-point hardware list; a brief description plus indicative runtime or memory usage would be more informative.
- Consider adding a brief rationale for the chosen network architecture (depth, width).
- In Table 1, adding "%" symbols could improve readability; please also clarify why the fraction of true samples decreases from BFB-I to BFB-II.
- Correct the typo "bach size" to "batch size."
- The introduction cites exclusively the authors’ previous work; including one or two references to the broader HEP context would improve balance.
Recommendation
Ask for minor revision