Gaia EDR3: Machine-Learned Radial Velocities for 91 Million Stars

Over 91 million predicted stellar radial velocities derived from Gaia EDR3 data using machine learning models. The catalog provides line-of-sight velocity estimates and uncertainties for a spatially complete sample of stars, created by researchers at Princeton University. A companion test set contains over 4.3 million stars for validation.

Use Cases

• Modeling galactic structure and dynamics based on predicted stellar velocities.
• Validating machine learning methods for astrophysical parameter estimation using the provided test set.
• Studying stellar populations and their kinematics across the Milky Way.
• Benchmarking new radial velocity prediction algorithms against this large-scale catalog.

Strengths

• Large scale with predictions for 91,840,346 stars.
• Includes a dedicated test set of 4,332,657 stars for validation.
• Provides both predicted velocity values and associated uncertainty estimates.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Gaia EDR3 data, processed by Adriana Dropulic at Princeton University.

Collection Method

Machine learning models applied to Gaia satellite data, as described in Dropulic et al. (arXiv:2205.12278).

Geography

Spatially complete sample of stars observed by the Gaia mission.