Galaxy Morphology Classification using Neural Ordinary Differential Equations

TL;DR

This paper presents a continuous-depth neural network model called NODE for galaxy morphology classification, achieving comparable accuracy to ResNet with fewer parameters, thus offering a more efficient alternative for large-scale astronomical image analysis.

Contribution

The paper introduces Neural Ordinary Differential Equations (NODE) for galaxy classification, demonstrating improved efficiency and comparable accuracy over traditional ResNet architectures.

Findings

NODE achieves 91-95% accuracy on Galaxy Zoo 2 dataset.

NODE uses about one-third the parameters of ResNet.

NODE overcomes ResNet's limitations in architecture tuning and data requirements.

Abstract

We introduce a continuous depth version of the Residual Network (ResNet) called Neural ordinary differential equations (NODE) for the purpose of galaxy morphology classification. We carry out a classification of galaxy images from the Galaxy Zoo 2 dataset, consisting of five distinct classes, and obtained an accuracy between 91-95\%, depending on the image class. We train NODE with different numerical techniques such as adjoint and Adaptive Checkpoint Adjoint (ACA) and compare them against ResNet. While ResNet has certain drawbacks, such as time consuming architecture selection (e.g. the number of layers) and the requirement of a large dataset needed for training, NODE can overcome these limitations. Through our results, we show that that the accuracy of NODE is comparable to ResNet, and the number of parameters used is about one-third as compared to ResNet, thus leading to a smaller memory footprint, which would benefit next generation surveys.