What does self-supervision add to engineered features for stellar light curves?
Abstract
TESS light curves encode a star's pulsation, rotation, binarity and oscillations, among other physical properties, but the survey produces far more light curves than can be labelled by hand. Astronomers have engineered a few dozen physically motivated features from them, which together remain a strong baseline. Previous work has trained encoders on light curves and judged them by their own downstream scores, but has not asked what they add to the engineered features. We pre-train a convolutional encoder on unlabelled TESS light curves to predict each window's latent representation from those of its neighbours. We then freeze it, add its features to the 25 engineered features, and read the combination out with one fixed linear probe on ten tasks: five variability classes, four asteroseismic targets and rotation period. Adding the learned features improves the readout on 7 of 10 tasks over the engineered features alone, where an untrained encoder does not. The combination also outperforms two networks trained end-to-end on the same labelled stars on most tasks. An encoder trained without labels therefore gives astronomers new features that complement the ones they already use rather than replace them.
Related
- Constraining the Light Curves of Magnified Stellar Events at z ~ 0.725
- Stellar flare detection and characterization in ZTF High-cadence light curves
- Stellar Flare Detection and Characterization in the ZTF High-cadence Light Curves
- Stellar Parameters and Magnetic Activity of Eruptive Variable Stars Based on TESS Light Curves and the LAMOST Spectral Survey
- An Irregular Shape of the Jupiter Trojan (3317) Paris Revealed by Stellar Occultations and Rotational Light Curves
BibTeX
@misc{liu2026what,
title = {{What does self-supervision add to engineered features for stellar light curves?}},
author = {Zijun Liu and Yue Ma and Christopher A. Theissen},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.14283},
eprint = {2610.14283},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.14283},
}
RIS
TY - UNPB
TI - What does self-supervision add to engineered features for stellar light curves?
AU - Zijun Liu
AU - Yue Ma
AU - Christopher A. Theissen
PY - 2026
DA - 2026/10/09
PB - arXiv
DO - 10.48550/arxiv.2610.14283
UR - https://arxiv.org/abs/2610.14283
AB - TESS light curves encode a star's pulsation, rotation, binarity and oscillations, among other physical properties, but the survey produces far more light curves than can be labelled by hand. Astronomers have engineered a few dozen physically motivated features from them, which together remain a strong baseline. Previous work has trained encoders on light curves and judged them by their own downstream scores, but has not asked what they add to the engineered features. We pre-train a convolutional encoder on unlabelled TESS light curves to predict each window's latent representation from those of its neighbours. We then freeze it, add its features to the 25 engineered features, and read the combination out with one fixed linear probe on ten tasks: five variability classes, four asteroseismic targets and rotation period. Adding the learned features improves the readout on 7 of 10 tasks over the engineered features alone, where an untrained encoder does not. The combination also outperforms two networks trained end-to-end on the same labelled stars on most tasks. An encoder trained without labels therefore gives astronomers new features that complement the ones they already use rather than replace them.
ER -