Multimodal Classification of Astronomical Transients

The Zwicky Transient Facility (ZTF) is capable of triggering hundreds of thousands of alerts a night for potential astronomical transients. Quick, accurate classification of these transients is critical for determining candidates for follow up observations. With the upcoming Legacy Survey of Space and Time from the Vera Rubin Observatory expecting to collect millions of astronomical transient alerts each night, rapid processing and classification is a pressing issue in time-domain astronomy. By using machine learning, the classification process can be automated, ensuring that interesting transients are identified and spectroscopically analyzed before fading. This multimodal model uses transformer architecture to classify astronomical transients from ZTF based on photometric time series data, image time series, and metadata.