当前位置:科学网首页 > 小柯机器人 >详情
科学家利用更多交互粒子变换器进行喷流标记
作者:小柯机器人 发布时间:2025/1/16 14:20:44

近日,上海理工大学理学院的王坤及其研究小组与河南大学物理与电子学院的朱经亚等人合作并取得一项新进展。经过不懈努力,他们利用更多交互粒子变换器进行喷流标记。相关研究成果已于2025年1月15日在《中国物理C》上发表。

本文介绍了更多交互粒子变换器(MIParT),这是一种为喷流标记设计的新型深度学习神经网络。该框架融入了该研究自主设计的更多交互注意力(MIA)机制,该机制增加了粒子交互嵌入的维度。研究人员使用顶夸克标记和夸克-胶子数据集对MIParT进行了测试。结果表明,MIParT不仅在准确率和AUC方面与LorentzNet及一系列洛伦兹等变方法相当,而且在背景抑制方面显著优于ParT模型。具体来说,在顶夸克标记数据集上,当信号效率为30%时,MIParT的背景抑制性能提高了约25%;在夸克-胶子数据集上,背景抑制性能提高了3%。

此外,MIParT仅需ParT所需参数的30%和计算复杂度的53%,证明了在降低模型复杂度的同时仍能实现高性能。对于非常大的数据集,研究人员将粒子嵌入的维度加倍,并将此变体称为MIParT-Large(MIParT-L)。他们发现,MIParT-L能进一步利用大型数据集中的知识。在1亿个JetClass数据集上预训练的模型基础上,微调后的MIParT-L在顶夸克标记数据集上的背景抑制性能提高了39%,在夸克-胶子数据集上提高了6%,超越了微调后的ParT。具体来说,与微调后的ParT相比,微调后的MIParT-L的背景抑制性能额外提高了2%。这些结果表明,MIParT有潜力提高粒子物理中喷流标记和事件识别的基准效率。

附:英文原文

Title: Jet tagging with more-interaction particle transformer

Author: Yifan Wu, Kun Wang, Congqiao Li, Huilin Qu, Jingya Zhu

Issue&Volume: 2025-01-15

Abstract: In this paper, we introduce the More-Interaction Particle Transformer (MIParT), a novel deep-learning neural network designed for jet tagging. This framework incorporates our own design, the More-Interaction Attention (MIA) mechanism, which increases the dimensionality of particle interaction embeddings. We tested MIParT using the top tagging and quark-gluon datasets. Our results show that MIParT not only matches the accuracy and AUC of LorentzNet and a series of Lorentz-equivariant methods, but also significantly outperforms the ParT model in background rejection. Specifically, it improves background rejection by approximately 25% with a signal efficiency of 30% on the top tagging dataset and by 3% on the quark-gluon dataset. Additionally, MIParT requires only 30% of the parameters and 53% of the computational complexity needed by ParT, proving that high performance can be achieved with reduced model complexity. For very large datasets, we double the dimension of particle embeddings, referring to this variant as MIParT-Large (MIParT-L). We found that MIParT-L can further capitalize on the knowledge from large datasets. From a model pre-trained on the 100M JetClass dataset, the background rejection performance of fine-tuned MIParT-L improves by 39% on the top tagging dataset and by 6% on the quark-gluon dataset, surpassing that of fine-tuned ParT. Specifically, the background rejection of fine-tuned MIParT-L improves by an additional 2% compared to that of fine-tuned ParT. These results suggest that MIParT has the potential to increase the efficiency of benchmarks for jet tagging and event identification in particle physics.

DOI: 10.1088/1674-1137/ad7f3d

Source: http://hepnp.ihep.ac.cn/article/doi/10.1088/1674-1137/ad7f3d

期刊信息

Chinese Physics C《中国物理C》,创刊于1977年。隶属于中国科学院高能物理研究所,最新IF:3.6

官方网址:http://hepnp.ihep.ac.cn/
投稿链接:https://mc03.manuscriptcentral.com/cpc


Baidu
map