ASTROMER A transformer-based embedding for the representation of light curves

Simple item page
dc.contributor.authorDonoso-Oliva, C.
dc.contributor.authorBecker, I.
dc.contributor.authorProtopapas, P.
dc.contributor.authorCabrera-Vives, G.
dc.contributor.authorVishnu, M.
dc.contributor.authorVardhan, H.
dc.date.accessioned2025-01-20T20:17:51Z
dc.date.available2025-01-20T20:17:51Z
dc.date.issued2023
dc.description.abstractTaking inspiration from natural language embeddings, we present ASTROMER, a transformer-based model to create representations of light curves. ASTROMER was pre-trained in a self-supervised manner, requiring no human-labeled data. We used millions of R-band light sequences to adjust the ASTROMER weights. The learned representation can be easily adapted to other surveys by re-training ASTROMER on new sources. The power of ASTROMER consists in using the representation to extract light curve embeddings that can enhance the training of other models, such as classifiers or regressors. As an example, we used ASTROMER embeddings to train two neural-based classifiers that use labeled variable stars from MACHO, OGLE-III, and ATLAS. In all experiments, ASTROMER-based classifiers outperformed a baseline recurrent neural network trained on light curves directly when limited labeled data were available. Furthermore, using ASTROMER embeddings decreases the computational resources needed while achieving state-of-the-art results. Finally, we provide a Python library that includes all the functionalities employed in this work.
dc.fuente.origenWOS
dc.identifier.doi10.1051/0004-6361/202243928
dc.identifier.eissn1432-0746
dc.identifier.issn0004-6361
dc.identifier.urihttps://doi.org/10.1051/0004-6361/202243928
dc.identifier.urihttps://repositorio.uc.cl/handle/11534/92424
dc.identifier.wosidWOS:000926204200008
dc.language.isoen
dc.revistaAstronomy & astrophysics
dc.rightsacceso restringido
dc.subjectmethods
dc.subjectstatistical - stars
dc.subjectstatistics - techniques
dc.subjectphotometric
dc.titleASTROMER A transformer-based embedding for the representation of light curves
dc.typeartículo
dc.volumen670
sipa.indexWOS
sipa.trazabilidadWOS;2025-01-12
Files
Collections
Artículos de revistas