SciPost Submission Page
Encoding off-shell effects in top pair production in Direct Diffusion networks
by Mathias Kuschick
This is not the latest submitted version.
Submission summary
| Authors (as registered SciPost users): | Mathias Kuschick |
| Submission information | |
|---|---|
| Preprint Link: | https://arxiv.org/abs/2412.17783v2 (pdf) |
| Date submitted: | Jan. 10, 2025, 8:33 p.m. |
| Submitted by: | Mathias Kuschick |
| Submitted to: | SciPost Physics Proceedings |
| Proceedings issue: | The 17th International Workshop on Top Quark Physics (TOP2024) |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approaches: | Computational, Phenomenological |
Abstract
To meet the precision targets of upcoming LHC runs in the simulation of top pair production events it is essential to also consider off-shell effects. Due to their great computational cost I propose to encode them in neural networks. For that I use a combination of neural networks that take events with approximate off-shell effects and transform them into events that match those obtained with full off-shell calculations. This was shown to work reliably and efficiently at leading order. Here I discuss first steps extending this method to include higher order effects.
Current status:
Reports on this Submission
Report #1 by Anonymous (Referee 1) on 2025-2-13 (Invited Report)
- Cite as: Anonymous, Report on arXiv:2412.17783v2, delivered 2025-02-13, doi: 10.21468/SciPost.Report.10665
Report
The writeup is a follow-up of a published paper, but it includes new results. I have only a few questions/requests: - the list of references is a little short and a little centered around the author group of the original paper. Please expand, including a comment on Schrodinger bridges, which are similar to direct diffusion. - reading the proceedings, I see that it goes beyond the original paper, but could the authors be more specific what is new and what changes under physics and technical aspects? - is the classifier also Bayesian? Would it make sense to train it that way or is the error from it sub-leading for instance in Fig.4.
Recommendation
Ask for minor revision