Journal article
Applying Multimodal Learning to Classify Transient Detections Early (AppleCiDEr). I. Dataset, Methods, and Infrastructure
Publications of the Astronomical Society of the Pacific, v 138(5), 54508
01 May 2026
Abstract
Modern time-domain surveys like the Zwicky Transient Facility (ZTF) and the Legacy Survey of Space and Time (LSST) generate hundreds of thousands to millions of alerts, demanding automatic, unified classification of transients and variable stars for efficient follow-up. We present Applying multimodal learning to Classify transient Detections Early ( AppleCiDEr) , a novel framework that integrates four key data modalities (photometry, image cutouts, metadata, and spectra) to overcome limitations of single-modality classification approaches. Our architecture introduces (i) two transformer encoders for photometry, (ii) a multimodal convolutional neural network (CNN) with domain-specialized metadata towers and Mixture-of-Experts fusion for combining metadata and images, and (iii) a CNN for spectra classification. Training on ∼30,000 real ZTF alerts, AppleCiDEr achieves high accuracy, allowing early identification and suggesting follow-up for rare transient spectra. The system provides the first unified framework for transients classification using real observational data, with seamless integration into brokering pipelines, demonstrating readiness for the LSST era.
Metrics
1 Record Views
Details
- Title
- Applying Multimodal Learning to Classify Transient Detections Early (AppleCiDEr). I. Dataset, Methods, and Infrastructure
- Creators
- Alexandra Junell - University of Minnesota SystemArgyro Sasli - University of Minnesota SystemFelipe Fontinele Nunes - University of Minnesota SystemMaojie Xu - University of Minnesota SystemBenny Border - University of Minnesota SystemNabeel Rehemtulla - Simons FoundationMariia Rizhko - University of California SystemYu-Jing Qin - California Institute of TechnologyTheophile Jegou Du LazAntoine Le Calloch - University of Minnesota SystemSushant Sharma Chaudhary - University of Minnesota SystemShaowei Wu - University of Minnesota SystemJesper Sollerman - Stockholm UniversityNiharika Sravan - Drexel UniversitySteven L. Groom - California Institute of TechnologyDavid Hale - California Institute of TechnologyMansi M. Kasliwal - California Institute of TechnologyJosiah Purdum - California Institute of TechnologyAvery Wold - California Institute of TechnologyMatthew J. Graham - California Institute of TechnologyMichael W. Coughlin - University of Minnesota System
- Publication Details
- Publications of the Astronomical Society of the Pacific, v 138(5), 54508
- Publisher
- IOP Publishing
- Number of pages
- 17
- Grant note
- National Science Foundation: PHY-2117997, PHY-2308862, PHY-2409481
The UMN authors acknowledge support from the National Science Foundation with grant Nos. PHY-2117997, PHY-2308862 and PHY-2409481.
- Resource Type
- Journal article
- Language
- English
- Academic Unit
- Physics
- Web of Science ID
- WOS:001773560900001
- Scopus ID
- 2-s2.0-105040010970
- Other Identifier
- 991022183056004721