Automated morphological classification of LMC-like galaxies through machine learning

Abstract

Understanding the morphological evolution of galaxies requires efficient and accu- rate classification methods, particularly in large-scale simulations and observations. This thesis presents a deep learning approach for the automated morphological classification of Large Magellanic Cloud-like galaxies, using high-resolution N-body simulations from the KRATOS suite. The study focuses on detecting structural features, such as bars and spiral arms, within the evolving stellar density maps of Large Magellanic Cloud-like galaxies undergoing gravitational interactions with the Milky Way and the Small Magellanic Cloud.

Using convolutional neural networks, the project developed binary classifiers for bar and spiral morphologies. The network was trained on 2D face-on galaxy density maps, ensuring alignment and centring based on stellar angular momentum. Despite challenges such as class imbalance and subtle morphological features, the network achieved 92% accuracy on unseen test data and generalised well to an unseen simulation (K20), where it maintained 89% classification accuracy.
Gradient-weighted Class Activation Mapping visualisations confirmed that the model consistently identified relevant morphological regions, while receiver operating characteristic and precision-recall curves indicated fair performance, more so in spiral classification.

The CNN model experienced limitations due to a small and imbalanced dataset, which could be addressed by using an expanded dataset and higher-resolution data. Finally, this work demonstrated that deep learning techniques can enhance the efficiency and consistency of galaxy morphology classification in simulations.

Popular Abstract

Our universe is home to billions of galaxies, each with a unique story, much like snowflakes, but on a cosmic scale. One of the Milky Way’s (MW) closest neighbours is the Large Magellanic Cloud (LMC), a smaller, irregularly shaped galaxy that orbits the MW. Yet, the LMC is not alone; it is in a gravitational-like dance with its smaller friend, the Small Magellanic Cloud (SMC). As these galaxies interact, their shapes warp, their stars shift, and they collide, giving a real-time lesson in how galaxies evolve.

Astronomers have long classified galaxies by painstakingly studying telescope images, a slow and subjective process. However, today, modern telescopes and advanced simulations generate too much data for humans to analyse manually. Imagine flipping through millions of photos one by one, trying to spot subtle differences. It is not just an annoying task, but also an impossible one! A better way to tackle these millions of photos is to use artificial intelligence (AI). Just as the phone camera can recognise faces in a crowd, AI can be trained to spot detailed patterns in galaxy shapes and sort them in seconds.

A modern simulation suite called KRATOS is used to test this. It recreates the gravitational interactions between the LMC, SMC, and the MW. The simulation tracks tens of millions of stars over cosmic time, predicting how their positions and movements change. However, raw data is not enough; it must be translated into something both humans and AI can understand.

First, the LMC stars are selected from the different particles in the KRATOS simulations. Then, star positions are turned into galaxy ”density maps.” 3D data is compressed into 2D maps, where brighter spots show densely packed stars, revealing the galaxy’s structure. Lastly, AI is taught to recognise galaxy shapes. Using a type of AI called Convolutional Neural Network (CNN), the same technology behind self-driving cars and medical imaging, it is trained to classify galaxies by their features. Over time, it learns to distinguish between spirals, barred galaxies, and other features.

This AI-powered approach can analyse galaxy data at incredible speed. Such models are already being applied to real galaxy observations from telescopes like the James Webb Space Telescope, helping the classification process and allowing astronomers to focus on interpretation and deeper analysis. The methods developed in this work could be used in future KRATOS simulations, showing how cosmic encounters drive the galaxies’ evolution. And since the LMC is on a collision course with the MW (do not worry, this will not happen for billions of years!), studying it now helps us predict what might happen when our galaxies finally merge.

• author: Mahmud, Marlin
• supervisor: Óscar Jiménez Arranz; Ross Church
• organization: Department of Physics; Astrophysics
• course: FYSK04 20251
• year: 2025
• type: M2 - Bachelor Degree
• subject: Physics and Astronomy
• report number: 2025-EXA245
• other publication id: 2025-EXA245
• language: English
• id: 9203178
• date added to LUP: 2025-06-19 14:24:16
• date last changed: 2025-06-19 14:24:16