标题： 使用机器学习预测晕形成时间 显示英文标题

标题： Predicting Halo Formation Time Using Machine Learning

摘要： 上下文：晕形成时间，它量化了暗物质晕的质量组装历史，直接影响星系的性质和演化。尽管无法直接观测，但可以通过恒星形成历史或星系空间分布等代理来推断。机器学习的最新进展使得利用星系和晕属性更准确地预测晕形成时间成为可能。目标：本研究旨在探讨一种基于机器学习的方法，利用来自宇宙学模拟的晕和重子属性来预测晕形成时间——定义为晕吸积其当前质量一半的时期。通过结合位于星系团中心的最亮星系团星系相关的属性、其相关的星系团内光成分和卫星星系，我们旨在超越这些分析预测，提高预测准确性并识别能够提供最佳晕组装历史代理的关键属性。方法：使用Three Hundred宇宙学模拟，我们使用随机森林（RF）和卷积神经网络（CNN）模型在晕和重子属性上进行训练，如质量、集中度、恒星质量和气体质量，以及最亮星系团星系和星系团内光的特征。CNN模型是在二维径向属性图上进行训练的。我们还仅使用可观察特征构建简单的线性模型。结果：RF模型的中位偏差为4%-9%，标准差为20%。CNN模型将中位偏差降低到<4%，尽管它们的散射更大。使用有限数量可观测值的简单线性模型实现了与RF模型相当的预测精度。传统的晕形成时间与质量/集中度之间的关系得到了保留。 显示英文摘要

摘要： Context:Halo formation time, which quantifies the mass assembly history of dark-matter halos, directly impacts galaxy properties and evolution. Although not directly observable, it can be inferred through proxies like star formation history or galaxy spatial distributions. Recent advances in machine learning enable more accurate predictions of halo formation time using galaxy and halo properties. Aims:This study aims to investigate a machine learning-based approach to predict halo formation time-defined as the epoch when a halo accretes half of its current mass-using both halo and baryonic properties derived from cosmological simulations. By incorporating properties associated with the brightest cluster galaxy located at the cluster center, its associated intracluster light component and satellite galaxies, we aim to surpass these analytical predictions, improve prediction accuracy and identify key properties that can provide the best proxy for the halo assembly history. Methods:Using The Three Hundred cosmological simulations, we train Random Forest (RF) and Convolutional Neural Network (CNN) models on halo and baryonic properties, such as mass, concentration, stellar and gas masses, and features of the brightest cluster galaxy and intracluster light. CNN models are trained on two-dimensional radial property maps. We also construct simple linear models using only observationally accessible features. Results:RF models show median biases of 4%-9% with standard deviations of 20%. CNN models reduce median bias to <4%, although they have higher scatter. Simple linear models using a limited number of observables achieve prediction accuracy comparable to RF models. Traditional relations between halo formation time and mass/concentration are preserved.