标题： 可扩展的多任务学习在异构图神经网络中的粒子对撞事件重建 显示英文标题

标题： Scalable Multi-Task Learning for Particle Collision Event Reconstruction with Heterogeneous Graph Neural Networks

摘要： 大型强子对撞机的亮度前沿不断增长，这对粒子碰撞事件的重建和分析提出了挑战。增加的粒子多重性正在考验数据获取阶段的延迟和存储需求，同时出现了新的复杂情况，包括更高的背景水平和更频繁的粒子顶点误关联。这反过来又需要开发更具整体性和可扩展性的重建方法，以利用机器学习领域的最新进展。 我们提出了一种新颖的异构图神经网络（HGNN）架构，该架构具有多样化的粒子碰撞关系的独特表示，并集成了图形剪枝层以实现可扩展性。 在一个模仿LHCb实验环境的环境中，采用多任务范式训练，这种HGNN显著提高了重味强子的重建性能。 值得注意的是，它在一个框架内同时执行粒子顶点关联和图形剪枝。 我们量化了重建和剪枝的性能，展示了推理时间随事件复杂性的增强可扩展性，并使用加权消息传递方案来缓解潜在的性能损失。 显示英文摘要

摘要： The growing luminosity frontier at the Large Hadron Collider is challenging the reconstruction and analysis of particle collision events. Increased particle multiplicities are straining latency and storage requirements at the data acquisition stage, while new complications are emerging, including higher background levels and more frequent particle vertex misassociations. This in turn necessitates the development of more holistic and scalable reconstruction methods that take advantage of recent advances in machine learning. We propose a novel Heterogeneous Graph Neural Network (HGNN) architecture featuring unique representations for diverse particle collision relationships and integrated graph pruning layers for scalability. Trained with a multi-task paradigm in an environment mimicking the LHCb experiment, this HGNN significantly improves beauty hadron reconstruction performance. Notably, it concurrently performs particle vertex association and graph pruning within a single framework. We quantify reconstruction and pruning performance, demonstrate enhanced inference time scaling with event complexity, and mitigate potential performance loss using a weighted message passing scheme.