ASTROCO: Self-Supervised Conformer-Style Transformers for Light-Curve Embeddings
This work addresses the challenge of label-efficient time-series analysis in astronomy, offering a domain-specific improvement for stellar light curve classification.
The paper tackled the problem of modeling irregular stellar light curves by introducing AstroCo, a Conformer-style encoder that combines attention with depthwise convolutions and gating, resulting in 70% and 61% lower error compared to Astromer v1 and v2, with a 7% relative macro-F1 gain.
We present AstroCo, a Conformer-style encoder for irregular stellar light curves. By combining attention with depthwise convolutions and gating, AstroCo captures both global dependencies and local features. On MACHO R-band, AstroCo outperforms Astromer v1 and v2, yielding 70 percent and 61 percent lower error respectively and a relative macro-F1 gain of about 7 percent, while producing embeddings that transfer effectively to few-shot classification. These results highlight AstroCo's potential as a strong and label-efficient foundation for time-domain astronomy.