Logo image
Applying Multimodal Learning to Classify Transient Detections Early (AppleCiDEr). I. Dataset, Methods, and Infrastructure
Journal article   Open access   Peer reviewed

Applying Multimodal Learning to Classify Transient Detections Early (AppleCiDEr). I. Dataset, Methods, and Infrastructure

Alexandra Junell, Argyro Sasli, Felipe Fontinele Nunes, Maojie Xu, Benny Border, Nabeel Rehemtulla, Mariia Rizhko, Yu-Jing Qin, Theophile Jegou Du Laz, Antoine Le Calloch, …
Publications of the Astronomical Society of the Pacific, v 138(5), 54508
01 May 2026
url
https://doi.org/10.1088/1538-3873/ae55cbView
Published, Version of Record (VoR) Open

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

Logo image