Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 23 min ago
PreprintCC BY 4.0arXiv

Learning Gaia: A Generative Model Trained on the Gaia Data Release 3 Source Catalog

  • John Soltis
Open at sourcePDFDOICite

Abstract

I train a conditional flow matching model on the positions, photometry, and astrometry of ∼ 579 million sources from the Gaia Data Release 3 Source Catalog. Using a probabilistic masking scheme during training, I develop a model capable of generating mock Gaia catalog sources with and without partial measurements of real Gaia sources. The model can produce coarse-grained mock Gaia Source Catalogs that reproduce the Milky Way disk and Magellanic Clouds, and can impute photometric and kinematic information. The model fails to reproduce small, distinct subpopulations like ω Centauri. This model is a necessary step towards a foundation model trained on the entirety of Gaia data products, and provides a blueprint for future work with the upcoming Gaia Data Release 4.

Related

  1. The Chandra-Gaia Catalog of Counterparts: Resolving ambiguous Gaia matches to X-ray sources in the Chandra Source Catalog using Machine Learning
    V. Samuel Pérez-Díaz, Vinay L. Kashyap, Joshua D. Ingram +6
    arXivPreprintCC BY 4.0Full textInstrumentation & methodsarXiv
  2. A cosmic filament catalog for COSMOS-Web Data Release 1 across 0.5 < z < 3.5
    Qiaoyang Hao, Tao Wang, Weishan Zhu +4
    arXivPreprintCC BY 4.0Full textCosmologyGalaxiesarXiv
  3. Recovering Missing Degrees of Freedom in Simulation-Trained Field-Level Generative Models
    Masato Shirasaki, Shiro Ikeda
    arXivPreprintCC BY 4.0Full textCosmologyarXiv
  4. Dark Matter Interpretations of the Galactic Center Gamma-Ray Excess with 18 Years of Fermi-LAT Data and the FL16Y Source Catalog
    Sang Hwan Kim, Hae Barg Kang, Seodong Shin
    arXivPreprintCC0Full textHigh-energy astrophysicsHigh-energy & particle physicsarXiv
  5. Image-catalog complementarity in quasar-galaxy ambiguous regimes of photometric source classification
    Ruqiao An, Chuan He, Yijun Pan +1
    Monthly Notices of the Royal Astronomical SocietyPeer-reviewedCC BY 4.0Full textSolar & stellar physicsCrossref

Cite

BibTeXRIS
BibTeX
@misc{soltis2026learning,
  title = {{Learning Gaia: A Generative Model Trained on the Gaia Data Release 3 Source Catalog}},
  author = {John Soltis},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.38326},
  eprint = {2609.38326},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.38326},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - Learning Gaia: A Generative Model Trained on the Gaia Data Release 3 Source Catalog
AU  - John Soltis
PY  - 2026
DA  - 2026/09/29
PB  - arXiv
DO  - 10.48550/arxiv.2609.38326
UR  - https://arxiv.org/abs/2609.38326
AB  - I train a conditional flow matching model on the positions, photometry, and astrometry of ∼ 579 million sources from the Gaia Data Release 3 Source Catalog. Using a probabilistic masking scheme during training, I develop a model capable of generating mock Gaia catalog sources with and without partial measurements of real Gaia sources. The model can produce coarse-grained mock Gaia Source Catalogs that reproduce the Milky Way disk and Magellanic Clouds, and can impute photometric and kinematic information. The model fails to reproduce small, distinct subpopulations like ω Centauri. This model is a necessary step towards a foundation model trained on the entirety of Gaia data products, and provides a blueprint for future work with the upcoming Gaia Data Release 4.
ER  -