Published on:
8 May 2024
Primary Category:
Earth and Planetary Astrophysics
Paper Authors:
Kathryn Volk,
Renu Malhotra
Model matches human judgement 98% of the time when classifying TNOs
Custom data features identify resonant dynamics missed by simpler methods
Vastly more efficient than manual classification with exponentially growing datasets
Will enable robust comparison of models to new observations from Vera Rubin Observatory
Provides classification probabilities; improves on binary secure/insecure designations
Machine learning classification of outer solar system objects
This paper presents a supervised machine learning model to classify icy bodies beyond Neptune called trans-Neptunian objects (TNOs) into different dynamical groups. The model was trained on a large and diverse dataset of real and synthetic TNO orbits. It uses custom features from numerical integrations to identify complex resonant dynamics. When tested, the model matched human judgement 98% of the time, offering major improvements in efficiency over manual classification as observations increase exponentially.
Using machine learning to classify astronomical transients
Automated detection of transients in images from the 4m ILMT
Automated discovery of unusual lunar surface features
Generation of Low-Inclination Neptune-Crossers by Planet Nine
Using deep learning to classify space objects in low Earth orbit
Forecasting abrupt cloud cover changes
No comments yet, be the first to start the conversation...
Sign up to comment on this paper