Szczegóły publikacji
Opis bibliograficzny
A modern machine learning approach for $B-meson$ decay generative modeling / Paweł KOPCIEWICZ, Michał Kacprzak, Tomasz SZUMLAK // Acta Physica Polonica. B, Proceedings Supplement ; ISSN 1899-2358. — 2023 — vol. 16 no. 3 art. no. 3-A5, s. 3-A5.1–3-A5.6. — Bibliogr. s. 3-A5.6, Abstr. — Publikacja dostępna online od: 2023-02-15. — XIV international conference on Beauty, Charm and Hyperon Hadrons : Kraków, Poland, 5–10 June, 2022
Autorzy (3)
Dane bibliometryczne
ID BaDAP | 146159 |
---|---|
Data dodania do BaDAP | 2023-04-11 |
Tekst źródłowy | URL |
DOI | 10.5506/APhysPolBSupp.16.3-A5 |
Rok publikacji | 2023 |
Typ publikacji | referat w czasopiśmie |
Otwarty dostęp | |
Creative Commons | |
Czasopismo/seria | Acta Physica Polonica, B, Proceedings Supplement |
Abstract
Recently, generative Machine Learning methods have gained a number of applications in physics analysis and detector science as they prove to be an excellent tool for data modeling. Generative Adversarial Networks (GANs), invented in 2014, were a breakthrough in computer vision and image synthesis, e.g. being able to generate realistic human faces. Although these networks show a growing presence in High-Energy Physics, their applications in this field are still quite scarce. The document investigates the potential use of GANs in Monte Carlo (MC) data generation in physics analysis, on the example of decay, simulated under the LHCb spectrometer geometry, and investigates whether they could be potentially used as an alternative method to the existing MC generators. Although GAN can quickly learn the shape of the distribution and recreate some correlations between physics parameters, its disadvantage is the lack of precision. The paper gives an overview of GAN capabilities as well as the optimized configuration of the model.