Representation Learning for Time-Domain High-Energy Astrophysics: Discovery of Extragalactic Fast X-ray Transient XRT 200515

2024-12-02

Monthly notices of the Royal Astronomical Society

AI Summary

To address the challenge of systematically detecting X-ray transient events (e.g., FXTs) in high-energy astrophysical time-domain data, this paper introduces a novel event representation paradigm integrating temporal and spectral information: constructing E–t images and E–t–dt cubes, and designing an end-to-end unsupervised pipeline comprising PCA or sparse autoencoder-based dimensionality reduction, DBSCAN clustering, and nearest-neighbor search. The method enables unified representation and analysis of variable-length X-ray event sequences for the first time. Applied to the Chandra archive, it identifies 3,559 transient candidates—including 3,447 flares and 112 dips—and discovers the first extragalactic fast X-ray transient, XRT 200515. This source exhibits a <10 s hard burst followed by an ~800 s oscillatory softening tail, potentially representing either a low-energy magnetar giant flare or a new class of Type-I X-ray bursts in the Large Magellanic Cloud.

Technology Category

• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Machine Learning: Time-Series/Data Streams
• Natural Language Processing: Information Extraction

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Topic discovery and tracking
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

We present a novel representation learning method for downstream tasks like anomaly detection or unsupervised classification in high-energy datasets. This enabled the discovery of a new extragalactic fast X-ray transient (FXT) in Chandra archival data, XRT 200515, a needle-in-the-haystack event and the first Chandra FXT of its kind. Recent serendipitous discoveries in X-ray astronomy, including FXTs from binary neutron star mergers and an extragalactic planetary transit candidate, highlight the need for systematic transient searches in X-ray archives. We introduce new event file representations, E − t Maps and E − t − dt Cubes, that effectively encode both temporal and spectral information, enabling the seamless application of machine learning to variable-length event file time series. Our unsupervised learning approach employs PCA or sparse autoencoders to extract low-dimensional, informative features from these data representations, followed by clustering in the embedding space with DBSCAN. New transients are identified within transient-dominant clusters or through nearest-neighbor searches around known transients, producing a catalog of 3,559 candidates (3,447 flares and 112 dips). XRT 200515 exhibits unique temporal and spectral variability, including an intense, hard <10 s initial burst, followed by spectral softening in an ∼800 s oscillating tail. We interpret XRT 200515 as either the first giant magnetar flare observed at low X-ray energies or the first extragalactic Type I X-ray burst from a faint, previously unknown LMXB in the LMC. Our method extends to datasets from other observatories such as XMM-Newton, Swift-XRT, eROSITA, Einstein Probe, and upcoming missions like AXIS.

Problem

• Develops representation learning for high-energy astrophysics data analysis.
• Discovers new extragalactic fast X-ray transient XRT 200515.
• Introduces E-t maps and E-t-dt cubes for machine learning applications.

Innovation

• Novel representation learning for high-energy data
• E-t maps and E-t-dt cubes encode temporal-spectral data
• Unsupervised learning with PCA and sparse autoencoders