Astra-CLR is a multi-filter time-series Transformer pre-trained via contrastive learning on 2.1 million ZTF light curves that achieves 0.70 accuracy classifying 12 variability classes, rising to 0.77 with partial fine-tuning.
An Improved Photometric Calibration of the Sloan Digital Sky Survey Imaging Data
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We present an algorithm to photometrically calibrate wide field optical imaging surveys, that simultaneously solves for the calibration parameters and relative stellar fluxes using overlapping observations. The algorithm decouples the problem of "relative" calibrations, from that of "absolute" calibrations; the absolute calibration is reduced to determining a few numbers for the entire survey. We pay special attention to the spatial structure of the calibration errors, allowing one to isolate particular error modes in downstream analyses. Applying this to the Sloan Digital Sky Survey imaging data, we achieve ~1% relative calibration errors across 8500 sq.deg. in griz; the errors are ~2% for the u band. These errors are dominated by unmodelled atmospheric variations at Apache Point Observatory. These calibrations, dubbed "ubercalibration", are now public with SDSS Data Release 6, and will be a part of subsequent SDSS data releases.
fields
astro-ph.IM 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Multi-Scale Contrastive Attention for Light-Curve Representation Learning
Astra-CLR is a multi-filter time-series Transformer pre-trained via contrastive learning on 2.1 million ZTF light curves that achieves 0.70 accuracy classifying 12 variability classes, rising to 0.77 with partial fine-tuning.