ttH(bb) dataset in the semi-leptonic decay channel

Higgs boson dataset in the \(t\bar{t}H(b\bar{b}) \) semil-leptonic channel, used for studies of deep learning and quantum machine learning classification studies [1, 2]. The simulation of the \(t\bar{t}H(b\bar{b}) \) semi-leptonic channel produces a data set that consists of the following features: Jet features: \((p_\mathrm{T}, \eta, \phi, E, \mathrm{b-tag}, p_\mathrm{x}, p_\mathrm{y}, p_\mathrm{z})\) Leptonic features: \((p_\mathrm{T}, \eta, \phi, E, p_\mathrm{x}, p_\mathrm{y}, p_\mathrm{z})\) Missing energy features: \((\phi, p_\mathrm{T}, p_\mathrm{x}, p_\mathrm{y})\) Before processing the data with (quantum) machine learning algorithms, the features are filtered using the following physically motivated criteria to constrain the problem in a suitable phase space. These criteria take into account the geometric acceptance of the detector and the goal of background suppression. The following preprocessing steps are applied in the related studies using this dataset: For electrons: \(p_\mathrm{T} > 30 \) GeV and \(|\eta|<2.1\) For muons: \(p_\mathrm{T} > 26\) GeV and \(|\eta|<2.1\) For jets: \(p_\mathrm{T} > 30\) GeV and \(|\eta|<2.4\) Isolation of the leptons with respect to jets is higher than the benchmark value of 0.1. Require at least 4 jets per event, at least 2 b-tagged jets, and exactly one lepton. The first seven most energetic jets are kept per collision event, allowing for one extra jet beyond the leading order expectation of 6 jets, to account for final state radiation. These criteria constrain the problem in a suitable phase space, taking into account the geometric acceptance of the CMS detector and the goal of background suppression. For more details, please see the corresponding papers. [1] V. Belis et al., Higgs analysis with quantum classifiers, EPJ Web Conf. 251, 03070 (2021), arXiv: 2104.07692. [2] V. Belis et al., Guided Quantum Compression for Higgs identification, arXiv: 2402.09524.本数据集为顶夸克-反顶夸克对伴随希格斯玻色子衰变至底夸克-反底夸克对（$tar{t}H(bar{b})$）半轻子道数据集，用于深度学习与量子机器学习分类研究[1,2]。 针对$tar{t}H(bar{b})$半轻子道的模拟实验生成了本数据集，其包含以下三类特征： 喷注（Jet）特征：$(p_mathrm{T}, eta, phi, E, ext{b-tag}, p_mathrm{x}, p_mathrm{y}, p_mathrm{z})$ 轻子特征：$(p_mathrm{T}, eta, phi, E, p_mathrm{x}, p_mathrm{y}, p_mathrm{z})$ 缺失能量特征：$(phi, p_mathrm{T}, p_mathrm{x}, p_mathrm{y})$ 在使用（量子）机器学习算法处理数据前，需基于物理动机的筛选准则将问题约束在合适的相空间内。这些准则兼顾了探测器的几何接收度与本底抑制的目标。 相关研究中针对本数据集采用了如下预处理流程： 针对电子：横向动量$p_mathrm{T} > 30$ GeV，快度$|eta| < 2.1$ 针对缪子：横向动量$p_mathrm{T} > 26$ GeV，快度$|eta| < 2.1$ 针对喷注：横向动量$p_mathrm{T} > 30$ GeV，快度$|eta| < 2.4$ 轻子相对于喷注的隔离度高于基准值0.1。 要求每个事件至少包含4个喷注、至少2个被标记的b喷注，且恰好存在一个轻子。 保留每次碰撞事件中前7个能量最高的喷注，相较于6喷注的领头阶预期额外允许1个喷注，以覆盖末态辐射效应。 这些准则将问题约束在合适的相空间内，兼顾了CMS探测器的几何接收度与本底抑制目标。更多细节请参阅对应论文。 [1] V. Belis等人，基于量子分类器的希格斯玻色子分析，EPJ Web Conf. 251, 03070 (2021), arXiv: 2104.07692. [2] V. Belis等人，用于希格斯玻色子识别的引导式量子压缩，arXiv: 2402.09524.