Large Radius Tracking Events in pp collisions with a Generic detector

This dataset is generated with the ACTS framework, using a generic detector description (similar to the one in the TrackML challenge) and a constant magnetic field of 2 Tesla in the beam direction (i.e. z-axis). The event simulation was done via the Fastras and the digitization was done via smearing functions, which is configured with the default configuration file in ACTS. We did not use an exact release when producing the dataset, but we believe release v15.0.0 should reproduce the dataset. We processed the original outputs from ACTS so that the dataset is machine-learning friendly. Four different physics processes are generated with the Pythia8 generator at 13 TeV of proton-proton collisions without the presence of addition collision (i.e. no pileup events). Heavy Neural Lepton with the neutral lepton mass of 15 GeV and a lifetime of 100 mm Heavy Neural Lepton with the neutral lepton mass of 10 GeV and a lifetime of 100 mm. Heavy Neural Lepton with the neutral lepton mass of 15 GeV and a lifetime of 200 mm. ttbar events The heavy neural lepton generated along with a neutrino through a virtual W boson is forced to decay to two muons and a muon neutrino. The dataset is saved as the torch dataset. One can access it via the following code snippet. <pre><code class="language-python">&gt;&gt;&gt; import torch &gt;&gt;&gt; data = torch.load("ttbar_PileUp000_10K_Processed/100") &gt;&gt;&gt; data Data(event_file="PileUp000_10K/event000000100", hid=[1634], layerless_true_edges=[2, 1477], layers=[1634], pid=[1634], x=[1634, 3])</code></pre> <em>x</em> is a table of spacepoints with dimension of [number of spacepoints, number of spacepoint features], the spacepoint features are their 3D global position. <em>pid</em> is a table of particle IDs each spacepoint associated with. hid is a table of spacepoint IDs, <em>layerless_true_edges</em> is a table of the true-level edges where edges connect spacepoints coming from the same track.本数据集基于ACTS框架（ACTS framework）生成，采用与TrackML挑战赛（TrackML challenge）中相似的通用探测器描述，并在束流方向（即z轴）施加2特斯拉的恒定磁场。事件模拟通过Fastras完成，数字化过程采用弥散函数（smearing functions），并使用ACTS中的默认配置文件进行配置。本次数据集生成未使用特定的精确发布版本，但我们认为v15.0.0版本应当可以复现该数据集。我们对ACTS的原始输出进行了后处理，以使数据集更适配机器学习任务。 本次数据集共包含四类不同的物理过程，均通过Pythia8生成器在13TeV质子-质子对撞条件下生成，且无额外碰撞（即无堆注事件）。四类过程分别为： 1. 中性轻子质量为15GeV、寿命为100mm的重中性轻子； 2. 中性轻子质量为10GeV、寿命为100mm的重中性轻子； 3. 中性轻子质量为15GeV、寿命为200mm的重中性轻子； 4. 顶夸克对事件（ttbar events）。 其中，通过虚W玻色子伴随中微子产生的重中性轻子，被强制衰变至两个μ子与一个μ子中微子。 本数据集以PyTorch数据集（torch dataset）格式存储，可通过以下代码片段访问： <pre><code class="language-python">>>> import torch >>> data = torch.load("ttbar_PileUp000_10K_Processed/100") >>> data Data(event_file="PileUp000_10K/event000000100", hid=[1634], layerless_true_edges=[2, 1477], layers=[1634], pid=[1634], x=[1634, 3])</code></pre> 其中，`x`为空间点（spacepoints）数据表，维度为`[空间点数量, 空间点特征数]`，空间点特征为其三维全局位置坐标。`pid`为每个空间点对应的粒子ID（particle ID）数据表。`hid`为空间点ID数据表，`layerless_true_edges`为真实层面边数据表，其中每条边连接来自同一径迹（track）的空间点。