Identifying Anomalous DESI Galaxy Spectra with a Variational Autoencoder

Journal Article · Fri Jun 20 04:00:00 EDT 2025 · No journal information

DOI:https://doi.org/10.48550/arXiv.2506.17376· OSTI ID:2572034

Nicolaou, C. ^[1]; Nathan, R. P. ^[1]; Lahav, O. ^[1]; Palmese, A. ^[2]; Saintonge, A. ^[1]; Aguilar, J. ^[3]; Ahlen, S. ^[4]; Prieto, C. Allende ^[5]; Bailey, S. ^[3]; BenZvi, S. ^[6]; Bianchi, D. ^[7]; Brodzeller, A. ^[3]; Brooks, D. ^[1]; Claybaugh, T. ^[3]; de la Macorra, A. ^[8]; Della Costa, J. ^[9]; Dey, Arjun ^[10]; Doel, P. ^[1]; Forero-Romero, J. E. ^[11]; Gaztañaga, E. ^[12] ; Gontcho A. Gontcho, S. ^[3]; Gutierrez, G. ^[13]; Honscheid, K. ^[14]; Howlett, C. ^[15]; Ishak, M. ^[16]; Kehoe, R. ^[17]; Kirkby, D. ^[18]; Kisner, T. ^[3]; Kremin, A. ^[3]; Lambert, A. ^[3]; Landriau, M. ^[3]; Le Guillou, L. ^[19]; Meisner, A. ^[10]; Miquel, R. ^[20]; Moustakas, J. ^[21]; Nadathur, S. ^[22]; Prada, F. ^[23]; Pérez-Ràfols, I. ^[24]; Rossi, G. ^[25]; Sanchez, E. ^[26]; Schubnell, M. ^[27]; Siudek, M. ^[28]; Sprayberry, D. ^[10]; Tarlé, G. ^[27]; Weaver, B. A. ^[10]; Zou, H. ^[29]

---

1. University Coll. London
2. Carnegie Mellon U.
3. LBL, Berkeley
4. Boston U.
5. Laguna U., Tenerife; IAC, La Laguna
6. Rochester U.
7. Milan U.; Brera Observ.
8. UNAM, Mexico
9. San Diego State U.; Natl. Solar Observ., Tucson
10. Natl. Solar Observ., Tucson
11. Andes U., Bogota
12. ICE, Bellaterra; Portsmouth U., ICG
13. Fermilab
14. Chicago U., KICP; Ohio State U.
15. Queensland U.
16. U. Texas, Dallas
17. Southern Methodist U.
18. UC, Irvine
19. LPNHE, Paris
20. ICREA, Barcelona; Barcelona, IFAE
21. Siena Coll., Loudonville
22. Portsmouth U., ICG
23. Granada U., Theor. Phys. Astrophys.
24. Barcelona, Polytechnic U.
25. Sejong U.
26. Madrid, CIEMAT
27. Michigan U.
28. IAC, La Laguna; Laguna U., Tenerife; ICE, Bellaterra
29. Beijing Observ.

The tens of millions of spectra being captured by the Dark Energy Spectroscopic Instrument (DESI) provide tremendous discovery potential. In this work we show how Machine Learning, in particular Variational Autoencoders (VAE), can detect anomalies in a sample of approximately 200,000 DESI spectra comprising galaxies, quasars and stars. We demonstrate that the VAE can compress the dimensionality of a spectrum by a factor of 100, while still retaining enough information to accurately reconstruct spectral features. We then detect anomalous spectra as those with high reconstruction error and those which are isolated in the VAE latent representation. The anomalies identified fall into two categories: spectra with artefacts and spectra with unique physical features. Awareness of the former can help to improve the DESI spectroscopic pipeline; whilst the latter can lead to the identification of new and unusual objects. To further curate the list of outliers, we use the Astronomaly package which employs Active Learning to provide personalised outlier recommendations for visual inspection. In this work we also explore the VAE latent space, finding that different object classes and subclasses are separated despite being unlabelled. We demonstrate the interpretability of this latent space by identifying tracks within it that correspond to various spectral characteristics. For example, we find tracks that correspond to increasing star formation and increase in broad emission lines along the Balmer series. In upcoming work we hope to apply the methods presented here to search for both systematics and astrophysically interesting objects in much larger datasets of DESI spectra.

• Research Organization: UNAM, Mexico; LPNHE, Paris; Queensland U.; ICREA, Barcelona; ICE, Bellaterra; Rochester U.; Chicago U., KICP; Siena Coll., Loudonville; University Coll. London; Sejong U.; Carnegie Mellon U.; Andes U., Bogota; U. Texas, Dallas; IAC, La Laguna; Portsmouth U., ICG; UC, Irvine; Granada U., Theor. Phys. Astrophys.; San Diego State U.; Boston U.; Brera Observ.; Barcelona, IFAE; Milan U.; Beijing Observ.; Natl. Solar Observ., Tucson; Madrid, CIEMAT; Ohio State U.; Michigan U.; Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); Barcelona, Polytechnic U.; Southern Methodist U.; Laguna U., Tenerife; Fermi National Accelerator Laboratory (FNAL), Batavia, IL (United States)
• Sponsoring Organization: US Department of Energy
• DOE Contract Number: 89243024CSC000002
• OSTI ID: 2572034
• Report Number(s): FERMILAB-PUB-25-0117-PPD; oai:inspirehep.net:2937997; arXiv:2506.17376
• Journal Information: No journal information, Journal Name: No journal information
• Country of Publication: United States
• Language: English