We present a simulation-based inference framework using a convolutional neural network to infer the dynamical mass of galaxy clusters from their observed 3D projected phase-space distribution, which consists of the projected galaxy positions in the sky and their line-of-sight velocities. By formulating the mass estimation problem within this simulation-based inference framework, we are able to quantify the uncertainties on the inferred masses in a straightforward and unbiased way. We generate a realistic mock catalogue emulating the Sloan Digital Sky Survey (SDSS) Legacy spectroscopic observations and explicitly illustrate the challenges posed by interloper (non-member) galaxies for cluster mass estimation from actual observations. Our approach constitutes the first optimal exploitation of the information content of the full projected phase-space distribution for the inference of dynamical cluster masses. We also present, for the first time, the application of a machine learning-based inference machinery to obtain dynamical masses of around $900$ galaxy clusters found in the SDSS Legacy Survey and demonstrate that the inferred masses reproduce the cluster mass function, as predicted by Planck $\Lambda$CDM cosmology, down to $10^{14.1}h^{-1}{\rm M}_{\odot}$ which is nearly a mass completeness limit of the selected cluster sample.

中文翻译：

---

使用 3D 卷积神经网络对动态星系团质量进行基于仿真的推断

我们提出了一个基于模拟的推理框架，使用卷积神经网络从观察到的 3D 投影相空间分布推断星系团的动力学质量，该分布由天空中投影的星系位置和它们的视线速度组成。通过在这个基于模拟的推理框架内制定质量估计问题，我们能够以直接和无偏见的方式量化推断质量的不确定性。我们生成了一个模拟斯隆数字巡天 (SDSS) 传统光谱观测的真实模拟目录，并明确说明了闯入者（非成员）星系对根据实际观测进行星团质量估计所带来的挑战。我们的方法构成了对全投影相空间分布的信息内容的首次优化利用，用于推断动态簇质量。我们还首次展示了基于机器学习的推理机制的应用，以获得在 SDSS Legacy Survey 中发现的约 900 美元星系团的动态质量，并证明推断的质量再现了星系团质量函数，如预测的那样普朗克 $\Lambda$CDM 宇宙学，低至 $10^{14.1}h^{-1}{\rm M}_{\odot}$，这几乎是所选集群样本的质量完整性极限。