A model local interpretation routine for deep learning based radio galaxy classification

TL;DR

This paper applies the LIME explainability technique to interpret deep learning models for radio galaxy classification, aiming to understand model predictions and compare them with astronomers' reasoning.

Contribution

It introduces the use of LIME for explaining deep learning predictions in radio galaxy classification, providing insights into model decision-making processes.

Findings

LIME helps interpret model predictions in radio galaxy classification.

Initial results show alignment between model explanations and astronomers' reasoning.

The approach offers a proof-of-concept for explainable AI in radio astronomy.

Abstract

Radio galaxy morphological classification is one of the critical steps when producing source catalogues for large-scale radio continuum surveys. While many recent studies attempted to classify source radio morphology from survey image data using deep learning algorithms (i.e., Convolutional Neural Networks), they concentrated on model robustness most time. It is unclear whether a model similarly makes predictions as radio astronomers did. In this work, we used Local Interpretable Model-agnostic Explanation (LIME), an state-of-the-art eXplainable Artificial Intelligence (XAI) technique to explain model prediction behaviour and thus examine the hypothesis in a proof-of-concept manner. In what follows, we describe how \textbf{LIME} generally works and early results about how it helped explain predictions of a radio galaxy classification model using this technique.