J/A+A/697/A107ivo://CDS.VizieR/J/A+A/697/A107doi:10.26093/cds/vizier.36970107CDSYe S.Cui W.Y.Li Y.B.Luo A.L.Jones H.R.A.2025-09-22T20:01:23Z2025-05-12T08:09:47ZCDS support team

CDS, Observatoire de Strasbourg, 11 rue de l'Universite, F-67000 Strasbourg, France

cds-question@unistra.frcarbon-starsvisible-astronomyA large fraction of Asymptotic Giant Branch (AGB) stars develop carbon-rich atmospheres during their evolution. Based on their color and luminosity, these carbon stars can be easily distinguished from many other kinds of stars. However, numerous G, K, and M giants also occupy the same region as carbon stars on the HR diagram. Despite this, their spectra exhibit differences, especially in the prominent CN molecular bands. We aim to distinguish carbon stars from other kinds of stars using Gaia's XP spectra, while providing attributional interpretations of key features necessary for identification, and even discovering additional new spectral key features. We propose a classification model named "GaiaNet", an improved one-dimensional convolutional neural network specifically designed for handling Gaia's XP spectra. We utilized the SHAP interpretability model to determine SHAP values for each feature in a spectrum, enabling us to explain the output of the "GaiaNet" model and provide further meaningful analysis. Compared to four traditional machine-learning methods, the "GaiaNet" model exhibits an average classification accuracy improvement of approximately 0.3% on the validation set, with the highest accuracy reaching 100%. Utilizing the SHAP model, we present a clear spectroscopic heatmap highlighting molecular band absorption features primarily distributed around CN773.3 and CN895.0, and summarize five key feature regions for carbon star identification. Upon applying the trained classification model to the CSTAR sample with Gaia "xp_sampled_mean" spectra, we obtained 451 new candidate carbon stars as a by-product. Our algorithm is capable of discerning subtle feature differences from low-resolution spectra of Gaia, thereby assisting us in effectively identifying carbon stars with typically higher temperatures and weaker CN features, while providing compelling attributive explanations. The interpretability analysis of deep learning holds significant potential in spectral identification.2025A&A...697A.107Yhttps://cdsarc.cds.unistra.fr/viz-bin/cat/J/A+A/697/A107CatalogResearchIsServedByTAP VizieR generic serviceIsServedByConesearch servicerelated-toI/355 : Gaia DR3 Part 1. Main source (Gaia Collaboration, 2022)https://cds.unistra.fr/vizier-org/licences_vizier.htmlhttps://vizier.cds.unistra.fr/viz-bin/VizieR-2?-source=J/A+A/697/A107https://vizier.iucaa.in/viz-bin/VizieR-2?-source=J/A+A/697/A107http://vizieridia.saao.ac.za/viz-bin/VizieR-2?-source=J/A+A/697/A107https://vizier.cds.unistra.fr/viz-bin/votable?-source=J/A+A/697/A107https://vizier.iucaa.in/viz-bin/votable?-source=J/A+A/697/A107http://vizieridia.saao.ac.za/viz-bin/votable?-source=J/A+A/697/A107GETtext/xml+votablehttps://tapvizier.cds.unistra.fr/TAPVizieR/tapCone search capability for table J/A+A/697/A107/tableb1 (Main information of carbon star candidates)https://vizier.cds.unistra.fr/viz-bin/conesearch/J/A+A/697/A107/tableb1?https://vizier.iucaa.in/viz-bin/conesearch/J/A+A/697/A107/tableb1?http://vizieridia.saao.ac.za/viz-bin/conesearch/J/A+A/697/A107/tableb1?GETtext/xml+votable180.050000true255.24443389-48.950651940.0055555555555555565/7205 7561 10552 6/1783 1961 2520 3187 3304 3459 3473 3590 3643 3654 3668-3669 3693-3694 3716 3740 4663 6305 12962 12969 12973 12986 13918 14074 14307-14308 14314 14343 14354 14356 14368 14402 14414 14418 14425-14426 14437 14449 14452 14545 14620 14624 14632 14703 15188 15378 15398 15447 15451 15454 15464 15476 15480 15483 15623 15639 15643-15644 15689 15719 21655 21897 23832 24475 24521 27978 28706-28707 28713-28714 28718 28729 28731 28735 28747 28778 28802 28806 28808-28809 28811 28817-28818 28825 28828-28829 28831 28834 28838-28839 28841-28842 28844-28846 28852-28853 28856 28858-28859 28861 28864 28866-28867 28872-28874 28876-28877 28882 28888 28890 28893 28895 28897-28898 28900 28904-28905 28909-28911 28915 28918-28919 28925 29066 29090 29093 29096 29099 29101 29107 29112 29114-29115 29162-29163 29167 29176 29191 29208 29249 29254 29258 29264 29445 29457 29508-29509 29511 29521-29524 29527-29529 29532-29533 29535 29588 30216-30217 30219 30222 30234-30235 30238 30242 30256 30258-30259 30262-30265 30269 30304 30306-30307 30312-30313 30316 30342 30351 30425 30448-30449 30526 30622 30632 30646-30647 30653 30658 30665 30670 30691 30696 30710 30715 30719 32085 37213 37339 37352 37365 37743 37752 37903 37907 37930-37931 38024 38528 39006 39239 39254 39399 39700 39820 39921 40483 40617 41465 41469 41591 41594-41596 41598 41615 41631 41653-41655 41658 41674-41676 41692 41696 41735 41757 41763-41764 41768-41769 41779 41800 41803 41808-41809 41813-41814 41819 41827-41828 41830 41856 42153-42156 42165-42166 42168 42187 42195 42199 42206 42213-42216 42218 42223-42225 42227 42229-42230 42233 42235 42298 42348 42351 42360 42367 42375-42376 42379-42381 42385 42390-42393 42400-42401 42404-42406 42410 42417 42433-42435 42437-42438 42453 42455 42458 42466 42493 42497 42562 42566 42968 43012-43013 43019-43020 43024 43042-43044 43048 43050https://cdsarc.cds.unistra.fr/viz-bin/moc/J/A+A/697/A107?format=asciiOpticaldefaultJ/A+A/697/A107/tableb1Main information of carbon star candidatesrecnoRecord number assigned by the VizieR team. Should Not be used for identification.meta.recordintGaiaDR3Gaia DR3 Source IDmeta.id;meta.mainRA_ICRSRight ascension (ICRS) at Ep=2016.0degpos.eq.ra;meta.mainDE_ICRSDeclination (ICRS) at Ep=2016.0degpos.eq.dec;meta.mainMainTypeMain type given by SIMBADsrc.classcharOtherTypesOther types given SIMBAD (2)src.classcharConfidenceConfidence given by GaiaNetstat.fit.paramdoubleSHAPvalueSum of SHAP values given by SHAP modelstat.fit.paramdoubleNotes[ab ] Note (1)meta.notechar