Star-Galaxy Image Separation with Computationally Efficient Gaussian Process Classification

Abstract

We introduce a novel method for discerning optical telescope images of stars from those of galaxies using Gaussian processes (GPs). Although applications of GPs often struggle in high-dimensional data modalities such as optical image classification, we show that a low-dimensional embedding of images into a metric space defined by the principal components of the data suffices to produce high-quality predictions from real large-scale survey data. We develop a novel method of GP classification hyperparameter training that scales approximately linearly in the number of image observations, which allows for application of GP models to large-size Hyper Suprime-Cam (HSC) Subaru Strategic Program data. In our experiments we evaluate the performance of a principal component analysis (PCA) embedded GP predictive model against other machine learning algorithms including a convolutional neural network and an image photometric morphology discriminator. Our analysis shows that our methods compare favorably with current methods in optical image classification while producing posterior distributions from the GP regression that can be used to quantify object classification uncertainty. We further describe how classification uncertainty can be used to efficiently parse large-scale survey imaging data to produce high-confidence object catalogs.

arXiv comment18 pages, 8 figures

References in corpus (30)

• SciPy 1.0--Fundamental Algorithms for Scientific Computing in Python
• LSST: from Science Drivers to Reference Design and Anticipated Data Products
• Large-Scale Galaxy Bias
• First Data Release of the Hyper Suprime-Cam Subaru Strategic Program
• Eight Ultra-faint Galaxy Candidates Discovered in Year Two of the Dark Energy Survey
• Cosmology with cosmic shear observations: a review
• The Hyper Suprime-Cam Software Pipeline
• The Dark Energy Survey Data Release 1
• Weak Gravitational Lensing with COSMOS: Galaxy Selection and Shape Measurements
• Dark Energy Survey Year 1 Results: Constraints on Extended Cosmological Models from Galaxy Clustering and Weak Lensing
• Star-galaxy Classification Using Deep Convolutional Neural Networks
• Primordial non-Gaussianity from the large scale structure
• Adversarial Examples, Uncertainty, and Transfer Testing Robustness in Gaussian Process Hybrid Deep Networks
• Measuring Dark Energy Clustering with CMB-Galaxy Correlations
• Characterization and Photometric Performance of the Hyper Suprime-Cam Software Pipeline
• Decision Tree Classifiers for Star/Galaxy Separation
• Excess Clustering on Large Scales in the MegaZ DR7 Photometric Redshift Survey
• Star-Galaxy Classification in Multi-Band Optical Imaging
• Relativistic Effect in Galaxy Clustering
• Euclid Mission: building of a Reference Survey
• A Hybrid Ensemble Learning Approach to Star-Galaxy Classification
• Angular correlation function of 1.5 million LRGs: clustering evolution and a search for BAO
• Gravity at the horizon: on relativistic effects, CMB-LSS correlations and ultra-large scales in Horndeski's theory
• Star-galaxy separation by far-infrared color-color diagrams for the AKARI FIS All-Sky Survey (Bright Source Catalogue Version beta-1)
• Star-galaxy classification in the Dark Energy Survey Y1 dataset
• A Fine-Grained Spectral Perspective on Neural Networks
• Effect of training characteristics on object classification: an application using Boosted Decision Trees
• Morphological Star-Galaxy Separation
• MuyGPs: Scalable Gaussian Process Hyperparameter Estimation Using Local Cross-Validation
• Star-Galaxy Separation via Gaussian Processes with Model Reduction

Cited by in corpus (2)

• Detection, Instance Segmentation, and Classification for Astronomical Surveys with Deep Learning (DeepDISC): Detectron2 Implementation and Demonstration with Hyper Suprime-Cam Data
• GAlaxy Light profile convolutional neural NETworks (GaLNets). I. fast and accurate structural parameters for billion galaxy samples