jet-universe/jetclass2

JetClass-II是一个大规模且全面的数据集，涵盖了广泛的大半径喷注特征和喷注的横向动量（pT）及质量值。该数据集旨在开发大型和全面的喷注模型，支持在大型强子对撞机（LHC）及其他场景中的广泛物理搜索和测量。数据集由三大部分组成：Res2P、Res34P和QCD，分别代表不同类型的喷注。每部分进一步细分为详细的类别，指示喷注的起源粒子。数据集还包括训练、验证和测试集，比例为4:1:1。数据集的结构包括喷注组成粒子、喷注本身、生成器级喷注组成粒子、生成器级喷注以及选定的真实粒子的特征。数据集生成使用了MadGraph + Pythia + Delphes，并模拟了LHC碰撞环境中的堆积效应。

JetClass-II is a large-scale and comprehensive dataset covering extensive large-radius jet signatures and a wide range of jet (p_mathrm{T}) and mass values. The dataset is designed to develop large and comprehensive jet models, intended for various applications, to support extensive physics searches and measurements at the Large Hadron Collider (LHC) and beyond. The dataset consists of three major parts: Res2P (generic (X) → 2 prong resonant jets), Res34P (generic (X) → 3 or 4 prong resonant jets), and QCD (jets from QCD multijet background). Each part is further subdivided into detailed categories, indicating which partons, leptons, or combinations thereof initiated the jet. The dataset also includes training, validation, and test sets, with a ratio of 4:1:1.

搜集汇总

背景与挑战

背景概述

JetClass-II是一个大规模高能物理数据集，专注于大半径喷注的分类和特性研究，包含约1.89亿行数据，涵盖Res2P、Res34P和QCD三大部分，用于训练喷注分类、回归和生成模型。数据集提供详细的粒子级、喷注级和生成级特征，支持物理搜索和测量应用，适用于机器学习和深度学习任务。