# The Cannon: A data-driven approach to stellar label determination

**Authors:** Melissa Ness, David W. Hogg, Hans-Walter Rix, Anna Y. Q. Ho, Gail Zasowski  
**arXiv:** [1501.07604](https://arxiv.org/abs/1501.07604)  
**Date:** 2015-01-29  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1501.07604v2/

## Key points

- The study discusses challenges in determining consistent stellar parameters and abundances for hundreds of thousands of stars in the Milky Way using spectroscopic surveys.
- The authors introduce a new data-driven approach called The Cannon, which learns from a subset of reference objects with known labels to determine stellar labels for the remaining survey stars.
- The model is trained on only 542 stars in 19 clusters as reference objects, with Teff, log g and [Fe/H] as the labels, and then applied to the spectra of 56,000 stars from APOGEE DR10.
- The Cannon performs well even at lower signal-to-noise levels and delivers comparably good labels even at one ninth the APOGEE observing time.
- Beyond the reference labels, The Cannon makes no use of stellar models nor any line-list but needs a set of reference objects that span label-space.
- The results show that The Cannon is very accurate, with its stellar labels comparing well to those provided by APOGEE pipeline (ASPCAP) labels in DR10, with rms differences that are basically identical to the stated ASPCAP uncertainties.
- The authors discuss the limitations of The Cannon and its future potential to bring different spectroscopic surveys onto a consistent scale of stellar labels.
- This approach could be extended to other astronomical applications such as galaxy classification or exoplanet characterization.

## Summary

The study discusses the challenges of determining consistent stellar parameters and abundances for hundreds of thousands of stars in the Milky Way using spectroscopic surveys. The authors introduce a new data-driven approach called The Cannon, which learns from a subset of reference objects with known labels to determine stellar labels for the remaining survey stars. The model is trained on only 542 stars in 19 clusters as reference objects, with Teff, log g and [Fe/H] as the labels, and then applied to the spectra of 56,000 stars from APOGEE DR10. The Cannon performs well even at lower signal-to-noise levels and delivers comparably good labels even at one ninth the APOGEE observing time. The study highlights that beyond the reference labels, The Cannon makes no use of stellar models nor any line-list but needs a set of reference objects that span label-space. The results show that The Cannon is very accurate, with its stellar labels comparing well to those provided by APOGEE pipeline (ASPCAP) labels in DR10, with rms differences that are basically identical to the stated ASPCAP uncertainties. The authors discuss the limitations of The Cannon and its future potential to bring different spectroscopic surveys onto a consistent scale of stellar labels. They suggest that this approach could be extended to other astronomical applications such as galaxy classification or exoplanet characterization. Overall, this study presents an innovative data-driven approach for determining stellar labels from spectroscopic data that has significant implications for understanding our galaxy's structure and evolution.