Research Paper#Astronomy, Machine Learning, Time Series Analysis🔬 ResearchAnalyzed: Jan 3, 2026 06:25
Transformer-based TDE Classifier for WFST
Published:Dec 31, 2025 11:02
•2 min read
•ArXiv
Analysis
This paper introduces a Transformer-based classifier, TTC, designed to identify Tidal Disruption Events (TDEs) from light curves, specifically for the Wide Field Survey Telescope (WFST). The key innovation is the use of a Transformer network ( exttt{Mgformer}) for classification, offering improved performance and flexibility compared to traditional parametric fitting methods. The system's ability to operate on real-time alert streams and archival data, coupled with its focus on faint and distant galaxies, makes it a valuable tool for astronomical research. The paper highlights the trade-off between performance and speed, allowing for adaptable deployment based on specific needs. The successful identification of known TDEs in ZTF data and the selection of potential candidates in WFST data demonstrate the system's practical utility.
Key Takeaways
- •Proposes a Transformer-based classifier (TTC) for identifying Tidal Disruption Events (TDEs) from light curves.
- •Utilizes a Transformer network ( exttt{Mgformer}) for improved performance and flexibility.
- •Designed for the Wide Field Survey Telescope (WFST) and can operate on real-time and archival data.
- •Demonstrates successful identification of known TDEs and selection of potential candidates.
- •Offers a trade-off between performance and speed through modular design.
Reference
“The exttt{Mgformer}-based module is superior in performance and flexibility. Its representative recall and precision values are 0.79 and 0.76, respectively, and can be modified by adjusting the threshold.”