Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bridging observations and simulations: a machine learning approach to galaxy clusters

Efrain Gatuzz

TL;DR

The paper tackles the challenge of comparing high-dimensional ICM LOS velocity maps from X-ray observations with theoretical predictions from simulations. It introduces a Siamese CNN trained with triplet loss to learn compact embeddings of velocity maps, enabling quantitative similarity assessment via Euclidean distance $d_E$ and cosine similarity $cos( obreak ullet, obreak ullet)$. The approach is validated on artificial perturbations of real XMM-Newton data, with embedding visualizations (e.g., t-SNE) showing coherent clustering around the originals, demonstrating robustness to plausible distortions. This framework provides a scalable, automated bridge between observations and simulations, with planned application to IllustrisTNG to evaluate how well current models reproduce observed ICM kinematics.

Abstract

The intracluster medium (ICM) records the history of galaxy clusters through its complex dynamical properties. To effectively interpret these properties, robust methods are needed to compare observational data with theoretical models. We present a novel machine learning framework for comparing ICM line-of-sight velocity maps derived from X-ray observations. Our approach uses convolutional and Siamese neural networks to identify similarities between different kinematic fields. We outline the architecture of this framework and perform a series of sanity checks to validate its performance. These checks demonstrate the model's ability to correctly identify and quantify kinematic features, establishing a powerful new tool for future comparative studies of the ICM.

Bridging observations and simulations: a machine learning approach to galaxy clusters

TL;DR

The paper tackles the challenge of comparing high-dimensional ICM LOS velocity maps from X-ray observations with theoretical predictions from simulations. It introduces a Siamese CNN trained with triplet loss to learn compact embeddings of velocity maps, enabling quantitative similarity assessment via Euclidean distance and cosine similarity . The approach is validated on artificial perturbations of real XMM-Newton data, with embedding visualizations (e.g., t-SNE) showing coherent clustering around the originals, demonstrating robustness to plausible distortions. This framework provides a scalable, automated bridge between observations and simulations, with planned application to IllustrisTNG to evaluate how well current models reproduce observed ICM kinematics.

Abstract

The intracluster medium (ICM) records the history of galaxy clusters through its complex dynamical properties. To effectively interpret these properties, robust methods are needed to compare observational data with theoretical models. We present a novel machine learning framework for comparing ICM line-of-sight velocity maps derived from X-ray observations. Our approach uses convolutional and Siamese neural networks to identify similarities between different kinematic fields. We outline the architecture of this framework and perform a series of sanity checks to validate its performance. These checks demonstrate the model's ability to correctly identify and quantify kinematic features, establishing a powerful new tool for future comparative studies of the ICM.
Paper Structure (5 sections, 2 figures)

This paper contains 5 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Galaxy Cluster Velocity Maps
  3. Machine Learning Methodology
  4. Validation Using Artificial Data
  5. Conclusion

Figures (2)

  • Figure 1: Examples from the artificial dataset generated for the Virgo cluster. These perturbed velocity maps were created by applying controlled transformations (e.g., rotation, noise) to the original XMM-Newton observation.
  • Figure 2: t-SNE projection of the embedding space for the Virgo cluster. The plot shows that the best matching correspond to the original observation, demonstrating that the Siamese CNN learns consistent kinematic representations.