On the performance of pre-trained vision transformers for supernova spectral classification using different spectral representations

概要

The spectroscopic classification of supernovae is a key component of time-domain astronomy and plays an important role in the identification of Type Ia events. The increasing volume and diversity of spectroscopic data motivate the development of automated classification approaches. In this work, we explore the use of Transformer-based vision models for supernova spectral classification, focusing on how different visual representations of spectra and fine-tuning strategies affect classification p…