Galaxy Morphology Classification with Counterfactual Explanation

2025-10-16

AI Summary

Deep learning models for galaxy morphology classification lack interpretability, hindering scientific trust and validation. Method: This paper proposes an invertible-flow-based encoder–decoder architecture that jointly optimizes supervised classification and differentiable counterfactual generation. By performing controlled, semantically meaningful perturbations in the latent space, the model synthesizes counterfactual galaxy images—e.g., altering spiral arm pitch or bar strength—while preserving structural fidelity. Contribution/Results: Our approach achieves state-of-the-art classification accuracy while enabling the first differentiable, causally traceable counterfactual explanations in astronomy. It quantitatively identifies how morphological features influence classification decisions, enhancing model transparency and domain expert trust. Experiments demonstrate robustness across diverse galaxy datasets and provide scientifically verifiable reasoning pathways, advancing the practical deployment of explainable AI in astronomical image analysis.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Galaxy morphologies play an essential role in the study of the evolution of galaxies. The determination of morphologies is laborious for a large amount of data giving rise to machine learning-based approaches. Unfortunately, most of these approaches offer no insight into how the model works and make the results difficult to understand and explain. We here propose to extend a classical encoder-decoder architecture with invertible flow, allowing us to not only obtain a good predictive performance but also provide additional information about the decision process with counterfactual explanations.

Problem

• Classifying galaxy morphologies using machine learning
• Explaining model decisions through counterfactual explanations
• Improving interpretability of automated galaxy classification systems

Innovation

• Encoder-decoder architecture extended with invertible flow
• Provides counterfactual explanations for decision process
• Achieves good predictive performance in classification