Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The weird and the wonderful in our Solar System: Searching for serendipity in the Legacy Survey of Space and Time

Brian Rogers, Chris J. Lintott, Steve Croft, Megan E. Schwamb, James R. A. Davenport

TL;DR

The paper addresses anomaly detection in LSST-era Solar System data by combining a deep autoencoder approach with latent-space similarity search to uncover objects that defy conventional categories, such as interstellar visitors. It demonstrates the autoencoder's ability to reveal interesting objects and uses two LSST-like simulations to validate the approach, while also benchmarking classic unsupervised detectors in a designed orbit-colour feature space. By constructing a Gaussian mixture model to generate synthetic global, cluster, and local anomalies and evaluating multiple detectors, the work highlights the strengths and limitations of different methods and emphasizes the need for task-specific tooling and feature expansion. Overall, the study provides a foundation for robust, scalable anomaly detection in large solar system catalogs and motivates future enhancements through richer feature spaces and ensemble strategies.

Abstract

We present a novel method for anomaly detection in Solar System object data, in preparation for the Legacy Survey of Space and Time. We train a deep autoencoder for anomaly detection and use the learned latent space to search for other interesting objects. We demonstrate the efficacy of the autoencoder approach by finding interesting examples, such as interstellar objects, and show that using the autoencoder, further examples of interesting classes can be found. We also investigate the limits of classic unsupervised approaches to anomaly detection through the generation of synthetic anomalies and evaluate the feasibility of using a supervised learning approach. Future work should consider expanding the feature space to increase the variety of anomalies that can be uncovered during the survey using an autoencoder.

The weird and the wonderful in our Solar System: Searching for serendipity in the Legacy Survey of Space and Time

TL;DR

The paper addresses anomaly detection in LSST-era Solar System data by combining a deep autoencoder approach with latent-space similarity search to uncover objects that defy conventional categories, such as interstellar visitors. It demonstrates the autoencoder's ability to reveal interesting objects and uses two LSST-like simulations to validate the approach, while also benchmarking classic unsupervised detectors in a designed orbit-colour feature space. By constructing a Gaussian mixture model to generate synthetic global, cluster, and local anomalies and evaluating multiple detectors, the work highlights the strengths and limitations of different methods and emphasizes the need for task-specific tooling and feature expansion. Overall, the study provides a foundation for robust, scalable anomaly detection in large solar system catalogs and motivates future enhancements through richer feature spaces and ensemble strategies.

Abstract

We present a novel method for anomaly detection in Solar System object data, in preparation for the Legacy Survey of Space and Time. We train a deep autoencoder for anomaly detection and use the learned latent space to search for other interesting objects. We demonstrate the efficacy of the autoencoder approach by finding interesting examples, such as interstellar objects, and show that using the autoencoder, further examples of interesting classes can be found. We also investigate the limits of classic unsupervised approaches to anomaly detection through the generation of synthetic anomalies and evaluate the feasibility of using a supervised learning approach. Future work should consider expanding the feature space to increase the variety of anomalies that can be uncovered during the survey using an autoencoder.
Paper Structure (7 sections, 4 equations, 2 figures)

This paper contains 7 sections, 4 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Data
  3. Classic approaches to anomaly detection.
  4. Methods
  5. Orbit-colour feature space
  6. Gaussian mixture modelling of the orbit-colour space.
  7. Unsupervised detection results

Figures (2)

  • Figure 1: Reconstruction of the orbit-colour space using the Gaussian mixture model.
  • Figure 2: The generated anomalies visualised using t-SNE. Global anomalies (red) are interwoven between the normal objects (blue). The clustered anomalies (orange) lie in dense groups around the edges of the normal objects in this reduced space. Local anomalies (black) tend to group round the edges of normal clusters of points.