Deep Learning Applications in Galaxy Morphology Classification

Deep learning approaches have revolutionised the classification of galaxy morphologies by automating the interpretation of vast imaging surveys. At its core, deep learning employs layered neural networks that learn hierarchical features directly from pixel data, eliminating the need for manual feature engineering. Such methods can distinguish between spheroidal, disk and irregular structures and can identify subtle features such as bars, spiral arms and merger signatures. Applications range from convolutional neural networks trained on millions of Sloan Digital Sky Survey images to transformer-based architectures capable of capturing both local texture and global context. By leveraging unsupervised clustering, autoencoding and supervised fine-tuning, these systems achieve high accuracy even in low signal-to-noise regimes and at high redshift. Advanced visualisation techniques, such as latent-space mapping with t-SNE, provide insight into the organisation of learned features and facilitate discovery of rare or transitional galaxies. The global significance of these developments lies in enabling consistent and scalable morphology catalogues, supporting studies of galaxy evolution, large-scale structure and the role of environment in shaping galactic form.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent studies have employed convolutional vision transformer models to refine five-class morphology tasks, demonstrating accuracies exceeding 98%. By integrating convolutional layers with transformer encoders, these models maintain sensitivity to local features while capturing long-range relationships, proving robust against image noise and variations in redshift.

Convolutional autoencoder frameworks have been deployed to build similarity-based retrieval engines, compressing galaxy images into compact latent representations. Such systems enable rapid identification of visually similar objects and achieve strong correlation with established morphology catalogues, accelerating the prioritisation of targets for detailed follow-up.

Hybrid clustering approaches combine unsupervised multiclustering with supervised deep networks to classify thousands of galaxies in infrared surveys. Initial clustering partitions galaxies into broad shape categories, followed by a fine-tuning stage using deep convolutional networks, ensuring complete classification of the sample and consistency with parametric and non-parametric morphological indicators.

Deep Learning Applications in Galaxy Morphology Classification publication trend

The graph below shows the total number of articles in deep learning applications in galaxy morphology classification across all publications each year (not limited to Nature Index journals).

Technical terms

Convolutional Neural Network (CNN): A deep learning model using convolutional layers to extract spatially localised features from images.

Convolutional Autoencoder (CAE): A neural network that compresses images into a latent space and reconstructs them, used for unsupervised feature learning.

Transformer: A deep learning architecture employing self-attention mechanisms to capture global dependencies within data.

Supervised Learning: A machine learning paradigm in which models are trained on labelled data to predict outcomes for new samples.

Unsupervised Learning: A machine learning approach that discovers patterns or groupings in unlabelled data.

Latent Space: A compressed representation of input data learned by models, where similar inputs are placed close together.

t-SNE (t-distributed Stochastic Neighbour Embedding): A dimensionality-reduction technique for visualising high-dimensional data in two or three dimensions.

References

1. The Classification of Galaxy Morphology in the H Band of the COSMOS-DASH Field: A Combination-based Machine-learning Clustering Model. The Astrophysical Journal Supplement Series (2023).
2. Similar Image Retrieval using Autoencoder. I. Automatic Morphology Classification of Galaxies. Publications of the Astronomical Society of the Pacific (2023).
3. Galaxy morphology classification based on Convolutional vision Transformer (CvT)★. Astronomy & Astrophysics (2024).
4. Improving galaxy morphologies for SDSS with Deep Learning. Monthly Notices of the Royal Astronomical Society (2018).