Spectral classification of brown dwarfs using machine learning

A.R. Callen; I.H. Bustos Fierro; M. G\'omez

arxiv: 2605.20146 · v1 · pith:GZSPOMXMnew · submitted 2026-05-19 · 🌌 astro-ph.SR · astro-ph.IM

Spectral classification of brown dwarfs using machine learning

A.R. Callen , I.H. Bustos Fierro , M. G\'omez This is my paper

Pith reviewed 2026-05-20 03:10 UTC · model grok-4.3

classification 🌌 astro-ph.SR astro-ph.IM

keywords brown dwarfsspectral classificationmachine learningphotometryRandom ForestGaussian Processes2MASSWISE

0 comments

The pith

Machine learning on 2MASS and WISE photometry classifies brown dwarf spectral types with F1 scores of 0.84 and 0.87.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper establishes that spectral types for brown dwarfs can be approximated using only photometric magnitudes from the 2MASS and WISE catalogs. The authors trained Random Forest and Gaussian Processes models on labeled data and validated them with cross-validation to obtain F1-scores of 0.84 and 0.87. They then used the models to classify 21 isolated brown dwarfs, finding most to be M6-M9 and a few L0-L4. A sympathetic reader would care because brown dwarfs require new spectral classes due to their cooling and the unexpected bluing effect makes visual classification difficult. This method could allow rapid classification for the growing number of discovered brown dwarfs.

Core claim

Random Forest and Gaussian Processes algorithms, trained with 5-fold cross-validation on 2MASS and WISE magnitudes, predict spectral types for brown dwarfs and achieve F1-scores of 0.84 and 0.87 on held-out test data. Application to 21 isolated brown dwarfs without previous classifications yields five objects between L0 and L4 and sixteen between M6 and M9.

What carries the argument

Random Forest and Gaussian Process classifiers that take multi-band photometric magnitudes from 2MASS and WISE as input features to predict discrete spectral type labels.

If this is right

The models offer a way to assign preliminary spectral types to large numbers of brown dwarfs using existing survey photometry.
Color information in the 2MASS and WISE bands captures patterns tied to the cooling sequence and bluing effect at the L/T transition.
This approach can prioritize objects for targeted spectroscopic follow-up.
Machine learning classification scales to future catalogs without requiring spectra for every candidate.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same models could be retrained on data from newer infrared surveys to classify brown dwarfs at greater distances.
Adding features such as proper motion might further reduce any residual effects from distance or reddening.
If the method works for M and L types, similar photometric training could be tested on T and Y dwarfs once more labeled examples become available.

Load-bearing premise

Photometric magnitudes from 2MASS and WISE alone contain enough information to distinguish spectral types without significant confusion from distance, reddening, or other astrophysical effects not captured in the training labels.

What would settle it

Spectroscopic observations that determine the actual spectral types of the 21 isolated brown dwarfs and compare them directly to the machine learning predictions would test whether the photometric classifications hold.

Figures

Figures reproduced from arXiv: 2605.20146 by A.R. Callen, I.H. Bustos Fierro, M. G\'omez.

**Figure 1.** Figure 1: Distribution of the 1723 isolated brown dwarfs in Galactic coordinates. The colour scale indicates the distance to each object in parsecs. Among the photometric data included in the catalogue, two of the most important sources are the 2MASS and WISE surveys, which provide near- and mid-infrared observations, respectively. 2MASS3 was a survey conducted between 1997 and 2001 that scanned approximately 70% of… view at source ↗

**Figure 2.** Figure 2: Top and middle panels: Colour–colour diagrams (J−K vs. J−H) as a function of spectral type. The spectral types highlighted in each panel are as follows: M0-M4 (a), M5-M9 (b), L0-L4 (c), L5-L9 (d), T0-T4 (e), T5-Y (f). The remaining objects in the sample are shown in grey for reference. Bottom panels: colour–magnitude diagrams, 𝑀𝐽 versus (𝐽 − 𝐻), (𝐻 − 𝐾𝑠 ), and (𝐽 − 𝐾𝑠 ), corresponding to panels (g), (h), a… view at source ↗

**Figure 3.** Figure 3: Colour-colour diagram (J−W1 vs. J−W2) for the brown dwarf sample, where points are colour-coded by spectral type. The black symbol with error bars in the lower-right corner represents the median uncertainty for all objects in the sample. Note: As indicated in [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 5.** Figure 5: The importance assigned by the Random Forest algorithm to each of the available features in the sample. an initial increasing trend, peaking around spectral types L5-T0, followed by a decrease toward later spectral types. In contrast, the colour H−K displays no variation across spectral types, thus suggesting W1−W2, J−H, and J−K as the most promising candidates. Subsequent classifier performance tests, bot… view at source ↗

**Figure 6.** Figure 6: 2MASS and WISE colour indices shown separately as a function of spectral type. Each of the indices has been vertically shifted for better visualization. Typical photometric uncertainties are on the order of ∼ 0.1 mag for the 2MASS colours and ∼ 0.02 mag for the WISE colour; these are not shown in the figure for clarity [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗

**Figure 7.** Figure 7: Combined 2MASS + WISE colour indices as a function of spectral type. Each of the indices has been vertically shifted for better visualization. Typical photometric uncertainties for all colours are on the order of ∼ 0.07 mag and are not shown in the figure for clarity. and (H−K) show a significant scatter of about 1 magnitude. Suárez et al. (2023) provide direct observational evidence that the atmospheres o… view at source ↗

**Figure 8.** Figure 8: Percentage of misclassified objects per spectral type range using the Random Forest (left) and Gaussian Processes (right) classifiers. In each panel, the solid line represents the average from 10 runs, while the shaded area indicates the standard deviation (blue for RF, pink for GP). employing distinct random states for the split of the sample into training/testing sets. This approach serves to eliminate a… view at source ↗

**Figure 9.** Figure 9: Confusion matrix between the spectral types predicted by RF and GP. The colour scale and the number in each matrix element represent the number of objects. 3.2. Comparison with previous studies In this section, we compare our spectral type determinations with results from three significant studies in the field to validate the robustness of our model. These studies utilized different methodologies for ident… view at source ↗

**Figure 10.** Figure 10: Confusion matrix comparing the spectral types predicted by our Random Forest model with those estimated by Skrzypek et al. (2015) using polynomial fits of colour indices. All comparisons are restricted to the subsets of sources from previous studies that met the required photometric features for our model; hence, these comparisons span narrower spectral ranges than those covered in the original studies. T… view at source ↗

**Figure 11.** Figure 11: Comparison of spectral types predicted by our Random Forest model and those reported in the literature. Left hand panel: Comparison with spectral types assigned by Skrzypek et al. (2015) (red) and Brooks et al. (2024) (blue). Darker colours indicate a higher number of objects per spectral subtype. Right hand panel: Spectral subtype estimates by Karpov et al. (2025) are shown as crosses for individual subt… view at source ↗

**Figure 12.** Figure 12: Confusion matrix comparing the literature spectral types and those predicted by the RF classifier for isolated brown dwarfs. The colour scale and the number in each matrix element represent the number of objects. 65%. For this reason, although both algorithms are generally consistent, when discrepancies arise, the probabilities provided by each should be examined. The spectral type with the highest probab… view at source ↗

**Figure 13.** Figure 13: Confusion matrix comparing the literature spectral types and those predicted by the GP classifier for isolated brown dwarfs. The colour scale and the number in each matrix element represent the number of objects [PITH_FULL_IMAGE:figures/full_fig_p010_13.png] view at source ↗

**Figure 14.** Figure 14: Confusion matrix comparing the spectral types predicted by RF and those predicted by the GP classifier for isolated brown dwarfs detected via direct imaging. The colour scale and the number in each matrix element represent the number of objects [PITH_FULL_IMAGE:figures/full_fig_p010_14.png] view at source ↗

read the original abstract

Brown dwarfs are compact objects that do not reach temperatures high enough to produce sustained hydrogen fusion. Consequently, they cool over time, gradually evolving through later spectral types. In fact, three new spectral types (L, T, and Y) were added to the Harvard sequence to accommodate the spectral features of brown dwarfs. During the cooling process, some brown dwarfs unexpectedly become bluer instead of redder (at optical and near-infrared wavelengths). This phenomenon, known as the bluing effect, is particularly noticeable at the L/T spectral transition. The aim of this work is to approximate the spectral type of brown dwarfs using only photometric data, in particular 2MASS and WISE magnitudes. We used two machine learning algorithms, Random Forest and Gaussian Processes, which were evaluated using a 70/30 train/test split. Both models were trained using 5-fold cross-validation and achieved F1-scores of 0.84 and 0.87, respectively, on the test set. After validating the reliability of the algorithms, we applied them to 21 isolated brown dwarfs without prior spectral type determinations. Our results indicate that 5 of these objects have a spectral type between L0 and L4, while the remaining 16 fall within the M6-M9 range. Machine learning algorithms, combined with multi-band photometry, are a powerful tool for estimating the spectral types of brown dwarfs.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Standard ML on 2MASS/WISE apparent magnitudes classifies brown dwarfs at F1 0.84-0.87 but risks picking up distance and reddening instead of intrinsic type.

read the letter

This paper trains Random Forest and Gaussian Process models on 2MASS and WISE photometry to estimate brown dwarf spectral types. It reports F1 scores of 0.84 and 0.87 on a 30 percent test split after 5-fold cross-validation, then assigns types to 21 new isolated objects, with five falling in L0-L4 and the rest in M6-M9. The concrete classifications for those 21 objects are the clearest new result, and the work shows a direct way to turn existing survey photometry into quick type estimates for large samples. That practical angle is the part that holds up. The setup follows routine supervised learning patterns already common in astronomy, so the contribution is incremental rather than a new method. The soft spot is the input choice. The models use apparent magnitudes, which bundle distance, extinction, and luminosity together with the temperature and atmospheric features that actually set spectral type. The abstract notes the bluing at the L/T transition, but if the training objects cover a range of distances or selection effects, the classifiers can latch onto those extra correlations. That makes the extension to the 21 new objects less certain without extra checks such as switching to colors or absolute magnitudes. Sample size, exact feature list, and class-balance handling are not described in the available text, which leaves the performance numbers only moderately supported. This is the sort of paper that would interest people running brown dwarf searches or mining wide-field catalogs for candidates. A reader who needs a photometric shortcut for triage would find it useful once the methods are filled in. It is worth sending to peer review so the authors can add the missing details and test the distance issue directly.

Referee Report

2 major / 2 minor

Summary. The manuscript applies Random Forest and Gaussian Process classifiers to predict spectral types of brown dwarfs from 2MASS and WISE apparent magnitudes. Models are trained with a 70/30 split and 5-fold cross-validation, yielding F1 scores of 0.84 and 0.87 on the test set. The classifiers are then applied to 21 isolated brown dwarfs without prior types, assigning 5 to L0-L4 and 16 to M6-M9.

Significance. A reliable photometry-only method for spectral typing would enable efficient classification of large brown-dwarf samples from wide-field surveys. The reported F1 scores suggest practical utility if the models capture intrinsic temperature and atmospheric features rather than distance or extinction correlations. Reproducibility would be aided by explicit sample sizes, feature lists, and code release, but the current presentation leaves the central performance claim only moderately supported.

major comments (2)

[Methods / Input features] The input features are apparent magnitudes from 2MASS and WISE (Abstract and Methods). Because spectral type is an intrinsic property tied to effective temperature, while apparent magnitudes include distance and reddening, the classifiers risk learning spurious correlations. The manuscript notes the bluing effect at the L/T transition but does not demonstrate that the models are insensitive to these extrinsic effects or compare against distance-independent colors or absolute magnitudes. This directly affects the reliability of the F1 scores and the classifications assigned to the 21 new objects.
[Results] No training-set size, number of objects per spectral class, feature list, or class-balance handling is reported (Abstract and Results). Without these details the F1 scores of 0.84 and 0.87 cannot be properly interpreted, and it is impossible to assess whether the 30 % test split is representative or whether performance is driven by a few dominant classes.

minor comments (2)

[Abstract] The abstract states both models were trained with 5-fold cross-validation yet evaluated on a 70/30 split; clarify whether the reported F1 scores are from the held-out test set after CV or from the CV folds themselves.
[Results] Error analysis, confusion matrices, or per-class performance metrics are absent; adding these would clarify where the models succeed or fail across the M-to-L range.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We are grateful to the referee for their thorough review and helpful suggestions. Below we provide point-by-point responses to the major comments. We have revised the manuscript accordingly to address the concerns raised.

read point-by-point responses

Referee: [Methods / Input features] The input features are apparent magnitudes from 2MASS and WISE (Abstract and Methods). Because spectral type is an intrinsic property tied to effective temperature, while apparent magnitudes include distance and reddening, the classifiers risk learning spurious correlations. The manuscript notes the bluing effect at the L/T transition but does not demonstrate that the models are insensitive to these extrinsic effects or compare against distance-independent colors or absolute magnitudes. This directly affects the reliability of the F1 scores and the classifications assigned to the 21 new objects.

Authors: We concur that apparent magnitudes are affected by distance and reddening, and that this could in principle lead to spurious correlations. However, the spectral type is strongly correlated with the spectral energy distribution, which is captured by the multi-band photometry even in apparent magnitudes for a sample with a range of distances. To directly address the referee's concern, in the revised version we will include an additional analysis using color indices (which are independent of distance) and compare the classification performance. We will also add a discussion on the potential effects of reddening and how the bluing effect at the L/T transition is accounted for in the model. This will provide stronger evidence that the classifiers are learning intrinsic properties. revision: yes
Referee: [Results] No training-set size, number of objects per spectral class, feature list, or class-balance handling is reported (Abstract and Results). Without these details the F1 scores of 0.84 and 0.87 cannot be properly interpreted, and it is impossible to assess whether the 30 % test split is representative or whether performance is driven by a few dominant classes.

Authors: We thank the referee for pointing out this oversight in the presentation. The revised manuscript will explicitly report the size of the training set, the number of objects in each spectral class, the complete list of input features (the specific 2MASS and WISE bands used), and any steps taken to handle class imbalance. These details will be added to the Methods section, allowing for a better assessment of the F1 scores and the validity of the train/test split. revision: yes

Circularity Check

0 steps flagged

Standard supervised ML validation on held-out labels shows no circularity

full rationale

The paper trains Random Forest and Gaussian Process models on 2MASS/WISE apparent magnitudes using known spectral-type labels from a training set, evaluates F1 scores on a 30% held-out test set after 5-fold cross-validation, and applies the fitted models to 21 new objects. This workflow uses external ground-truth labels for both training and test evaluation; the reported performance metrics are measured against independent test labels rather than being re-derived from the model parameters or training inputs by construction. No self-citations, uniqueness theorems, or ansatzes are invoked to justify the core results, and the classification of new objects is a forward application rather than a tautological restatement of fitted quantities. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the quality and representativeness of the labeled training set and the assumption that broadband photometry encodes spectral-type information with limited contamination.

free parameters (1)

Random Forest and Gaussian Process hyperparameters
Number of trees, kernel choices, and regularization parameters are fitted during training but not reported in the abstract.

axioms (1)

domain assumption The training sample of brown dwarfs with known spectral types is representative of the objects to which the models are applied.
Supervised learning requires this for generalization to the 21 new targets.

pith-pipeline@v0.9.0 · 5785 in / 1296 out tokens · 49076 ms · 2026-05-20T03:10:40.981550+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We used two machine learning algorithms, Random Forest and Gaussian Processes... achieved F1-scores of 0.84 and 0.87... applied them to 21 isolated brown dwarfs
IndisputableMonolith/Foundation/AlphaCoordinateFixation.lean J_uniquely_calibrated_via_higher_derivative unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

selected features were the absolute magnitudes J, H, and W1, and the colours W1-W2, J-H, J-W1, J-W2

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

[1]

Avdeeva,A.2023,AstronomyandComputing,45,100744

work page 2023
[2]

L., Jones, H

Baig, S., Smart, R. L., Jones, H. R. A., Gagné, J., Pinfield, D. J., Cheng,G.,&Moranta,L.2024,MNRAS,533,3784

work page 2024
[3]

2014, Publication of the Astronomical Society of Australia,31,e043 Baraffe,I.,Homeier,D.,Allard,F.,&Chabrier,G.2015,A&A,577, A42

Bailey, J. 2014, Publication of the Astronomical Society of Australia,31,e043 Baraffe,I.,Homeier,D.,Allard,F.,&Chabrier,G.2015,A&A,577, A42

work page 2014
[4]

Bayo, A., Rodrigo, C., Barrado Y Navascués, D., Solano, E., Gutiérrez,R.,Morales-Calderón,M.,&Allard,F.2008,A&A, 492,277 Beiler,S.A.,Cushing,M.C.,Kirkpatrick,J.D.,Schneider,A.C., Mukherjee,S.,Marley,M.S.,Marocco,F.,&Smart,R.L.2024, ApJ,973,107 Bessel,M.S.1990,A&AS,83,357 Bessell,M.&Murphy,S.2012,PASP,124,140

work page 2008
[5]

Huldtgren, M., Abergel, A., Blommaert, J., Boulanger, F., Burgdorf,M.,Cesarsky,C.J.,Cesarsky,D.,Copet,E.,Davies,J., Falgarone,E.,Lagache,G.,Montmerle,T.,Pérault,M.,Persi,P., Prusti,T.,Puget,J.L.,&Sibille,F.2001,A&A,372,173 Boone,K.2019,AJ,158,257 Breiman,L.2001,MachineLearning,45,5

work page 2001
[6]

Mod.Phys.,73,719[LINK]

Ruiz, L., Sanchez, J., Sainio, A., Schümann, J., Schonau, M., Tanner,C.,Stevnbak,N.,Stenner,A.,Thévenot,M.,Thakur,V., Voloshin,N.V.,W e ¸dracki,Z.,&TheBackyardWorlds: Planet 9Collaboration.2024,AJ,168,211 Burhanudin,U.F.&Maund,J.R.2023,MNRAS,521,1601 Burrows,A.,Hubbard,W.B.,Lunine,J.I.,&Liebert,J.2001,Rev. Mod.Phys.,73,719[LINK]

work page 2024
[7]

& Liebert, J

Burrows, A. & Liebert, J. 1993, Reviews of Modern Physics, 65, 301 Caballero,J.A.2008,A&A,478,667

work page 1993
[8]

2019,A&A,627,A167

Calissendorff, P., Janson, M., Asensio-Torres, R., & Köhler, R. 2019,A&A,627,A167

work page 2019
[9]

B., Henry, T

Dieterich, S. B., Henry, T. J., Jao, W.-C., Winters, J. G., Hosey, A.D.,Riedel,A.R.,&Subasavage,J.P.2014,AJ,147,94 Dransfield,G.&Triaud,A.H.M.J.2020,MNRAS,499,505

work page 2014
[10]

K., Cruz, K

Nienartowicz, K., Jan, A., Beck, M., Mowlavi, N., De Cat, P., Lebzelter,T.,&Eyer,L.2011,MNRAS,414,2602 Filippazzo,J.C.,Rice,E.L.,Faherty,J.,Cruz,K.L.,VanGordon, M.M.,&Looper,D.L.2015,ApJ,810,158 Gagné, J., Faherty, J. K., Cruz, K. L., Lafreniére, D., Doyon, R., Malo,L.,Burgasser,A.J.,Naud,M.-E.,Artigau,É.,Bouchard, S.,Gizis,J.E.,&Albert,L.2015,ApJS,219,33...

work page 2011
[11]

E., Yoachim, P., Jones, R

Gizis, J. E., Yoachim, P., Jones, R. L., Hilligoss, D., & Ji, J. 2022, ApJS,263,23

work page 2022
[12]

Greiner, J., Bolmer, J., Yates, R. M., Habouzit, M., Bañados, E., Afonso,P.M.J.,&Schady,P.2021,A&A,654,A79 Holwerda,B.,Pirzkal,N.,Burgasser,A.,&Hsu,C.-C.2023,arXiv e-prints,arXiv:2306.12363 Itoh,Y.,Tamura,M.,&Nakajima,T.1999,AJ,117,1471 Kang,G.,Lim,J.,&Seo,B.2024,arXive-prints,arXiv:2409.04490 Karpov,S.,Malkov,O.,&Avdeeva,A.2025,A&A,695,A195 Kirkpatrick,J...

work page arXiv 2021
[13]

F., Marsh, K

Skrutskie, M. F., Marsh, K. A., Wright, E. L., Mainzer, A., Eisenhardt,P.R.,McLean,I.S.,Thompson,M.A.,Bauer,J.M., Benford,D.J.,Bridge,C.R.,Lake,S.E.,Petty,S.M.,Stanford, S. A., Tsai, C.-W., Bailey, V., Beichman, C. A., Bloom, J. S., 40 Callen et al. Spectral classification of brown dwarfs using machine learning Bochanski,J.J.,Burgasser,A.J.,Capak,P.L.,Cru...

work page 2011
[14]

Gonzales, E. C., Hsu, C.-C., Kamraj, N., Kiman, R., Rees, J., Theissen,C.,Ammar,K.,Andersen,N.S.,Beaulieu,P.,Colin, G.,Elachi,C.A.,Goodman,S.J.,Gramaize,L.,Hamlet,L.K., Hong,J.,Jonkeren,A.,Khalil,M.,Martin,D.W.,Pendrill,W., Pumphrey,B.,Rothermich,A.,Sainio,A.,Stenner,A.,Tanner, C., Thévenot, M., Voloshin, N. V., Walla, J., Wędracki, Z., & BackyardWorlds: ...

work page 2021
[15]

P., Smith, J

Schneider, D. P., Smith, J. A., Strauss, M. A., Uomoto, A., & York,D.G.2002,ApJ,564,452 Leggett,S.K.,Tremblin,P.,Esplin,T.L.,Luhman,K.L.,&Morley, C.V.2017,ApJ,842,118 Marley,M.S.,Saumon,D.,Visscher,C.,Lupu,R.,Freedman,R., Morley,C.,Fortney,J.J.,Seay,C.,Smith,A.J.R.W.,Teal,D.J., &Wang,R.2021,ApJ,920,85 Meisner,A.M.,Caselden,D.,Schlafly,E.F.,&Kiwy,F.2023,AJ...

work page 2002
[16]

Monin, J. L., Guieu, S., Pinte, C., Rebull, L., Goldsmith, P., Fukagawa,M.,Ménard,F.,Padgett,D.,Stappelfeld,K.,McCabe, C.,Carey,S.,Noriega-Crespo,A.,Brooke,T.,Huard,T.,Terebey, S.,Hillenbrand,L.,&Guedel,M.2010,A&A,515,A91 Nakajima,T.,Oppenheimer,B.R.,Kulkarni,S.R.,Golimowski, D.A.,Matthews,K.,&Durrance,S.T.1995,Nature,378,463 Pedregosa,F.,Varoquaux,G.,Gra...

work page 2010
[17]

M., Drummond, B., & Hébrard, E

Spiegelman, F., Goyal, J. M., Drummond, B., & Hébrard, E. 2020,A&A,637,A38 Plewa,P.M.2018,MNRAS,476,3974 Rilinger,A.M.&Espaillat,C.C.2021,ApJ,921,182 Saumon,D.&Marley,M.S.2008,ApJ,689,1327 Schneider,A.C.,Munn,J.A.,Vrba,F.J.,Bruursema,J.,Dahm, S.E.,Williams,S.J.,Liu,M.C.,&Dorland,B.N.2023,AJ,166, 103

work page 2020
[18]

M., Beichman, C., Capps, R., Chester,T.,Elias,J.,Huchra,J.,Liebert,J.,Lonsdale,C.,Monet, D.G.,Price,S.,Seitzer,P.,Jarrett,T.,Kirkpatrick,J.D.,Gizis, J

Schneider, S., Carpenter, J. M., Beichman, C., Capps, R., Chester,T.,Elias,J.,Huchra,J.,Liebert,J.,Lonsdale,C.,Monet, D.G.,Price,S.,Seitzer,P.,Jarrett,T.,Kirkpatrick,J.D.,Gizis, J. E., Howard, E., Evans, T., Fowler, J., Fullmer, L., Hurt, R., Light,R.,Kopan,E.L.,Marsh,K.A.,McCallon,H.L.,Tam,R., VanDyk,S.,&Wheelock,S.2006,AJ,131,1163

work page 2006
[19]

J., Faherty, J

Skrzypek, N., Warren, S. J., Faherty, J. K., Mortlock, D. J., Burgasser,A.J.,&Hewett,P.C.2015,A&A,574,A78 Suárez,G.&Metchev,S.2022,MNRAS,513,5701 Suárez,G.,Vos,J.M.,Metchev,S.,Faherty,J.K.,&Cruz,K.2023, ApJ,954,L6 Tarter,J.C.1975,PhDthesis,UniversityofCalifornia,Berkeley Testi,L.,Natta,A.,Scholz,A.,Tazzari,M.,Ricci,L.,&deGregorio Monsalvo,I.2016,A&A,593,A...

work page 2015
[20]

J., Blain, A., Mendez, B., Irace, W

McLean, I., Benford, D., Lonsdale, C. J., Blain, A., Mendez, B., Irace, W. R., Duval, V., Liu, F., Royer, D., Heinrichsen, I., Howard,J.,Shannon,M.,Kendall,M.,Walsh,A.L.,Larsen,M., Cardon,J.G.,Schick,S.,Schwalm,M.,Abid,M.,Fabinsky,B., Naes,L.,&Tsai,C.-W.2010,AJ,140,1868 Ye,G.,Zhang,H.,&Wu,Q.2024,ApJS,275,19 APPENDICES A. Appendix: Spectral types predicted...

work page 2010

[1] [1]

Avdeeva,A.2023,AstronomyandComputing,45,100744

work page 2023

[2] [2]

L., Jones, H

Baig, S., Smart, R. L., Jones, H. R. A., Gagné, J., Pinfield, D. J., Cheng,G.,&Moranta,L.2024,MNRAS,533,3784

work page 2024

[3] [3]

2014, Publication of the Astronomical Society of Australia,31,e043 Baraffe,I.,Homeier,D.,Allard,F.,&Chabrier,G.2015,A&A,577, A42

Bailey, J. 2014, Publication of the Astronomical Society of Australia,31,e043 Baraffe,I.,Homeier,D.,Allard,F.,&Chabrier,G.2015,A&A,577, A42

work page 2014

[4] [4]

Bayo, A., Rodrigo, C., Barrado Y Navascués, D., Solano, E., Gutiérrez,R.,Morales-Calderón,M.,&Allard,F.2008,A&A, 492,277 Beiler,S.A.,Cushing,M.C.,Kirkpatrick,J.D.,Schneider,A.C., Mukherjee,S.,Marley,M.S.,Marocco,F.,&Smart,R.L.2024, ApJ,973,107 Bessel,M.S.1990,A&AS,83,357 Bessell,M.&Murphy,S.2012,PASP,124,140

work page 2008

[5] [5]

Huldtgren, M., Abergel, A., Blommaert, J., Boulanger, F., Burgdorf,M.,Cesarsky,C.J.,Cesarsky,D.,Copet,E.,Davies,J., Falgarone,E.,Lagache,G.,Montmerle,T.,Pérault,M.,Persi,P., Prusti,T.,Puget,J.L.,&Sibille,F.2001,A&A,372,173 Boone,K.2019,AJ,158,257 Breiman,L.2001,MachineLearning,45,5

work page 2001

[6] [6]

Mod.Phys.,73,719[LINK]

Ruiz, L., Sanchez, J., Sainio, A., Schümann, J., Schonau, M., Tanner,C.,Stevnbak,N.,Stenner,A.,Thévenot,M.,Thakur,V., Voloshin,N.V.,W e ¸dracki,Z.,&TheBackyardWorlds: Planet 9Collaboration.2024,AJ,168,211 Burhanudin,U.F.&Maund,J.R.2023,MNRAS,521,1601 Burrows,A.,Hubbard,W.B.,Lunine,J.I.,&Liebert,J.2001,Rev. Mod.Phys.,73,719[LINK]

work page 2024

[7] [7]

& Liebert, J

Burrows, A. & Liebert, J. 1993, Reviews of Modern Physics, 65, 301 Caballero,J.A.2008,A&A,478,667

work page 1993

[8] [8]

2019,A&A,627,A167

Calissendorff, P., Janson, M., Asensio-Torres, R., & Köhler, R. 2019,A&A,627,A167

work page 2019

[9] [9]

B., Henry, T

Dieterich, S. B., Henry, T. J., Jao, W.-C., Winters, J. G., Hosey, A.D.,Riedel,A.R.,&Subasavage,J.P.2014,AJ,147,94 Dransfield,G.&Triaud,A.H.M.J.2020,MNRAS,499,505

work page 2014

[10] [10]

K., Cruz, K

Nienartowicz, K., Jan, A., Beck, M., Mowlavi, N., De Cat, P., Lebzelter,T.,&Eyer,L.2011,MNRAS,414,2602 Filippazzo,J.C.,Rice,E.L.,Faherty,J.,Cruz,K.L.,VanGordon, M.M.,&Looper,D.L.2015,ApJ,810,158 Gagné, J., Faherty, J. K., Cruz, K. L., Lafreniére, D., Doyon, R., Malo,L.,Burgasser,A.J.,Naud,M.-E.,Artigau,É.,Bouchard, S.,Gizis,J.E.,&Albert,L.2015,ApJS,219,33...

work page 2011

[11] [11]

E., Yoachim, P., Jones, R

Gizis, J. E., Yoachim, P., Jones, R. L., Hilligoss, D., & Ji, J. 2022, ApJS,263,23

work page 2022

[12] [12]

Greiner, J., Bolmer, J., Yates, R. M., Habouzit, M., Bañados, E., Afonso,P.M.J.,&Schady,P.2021,A&A,654,A79 Holwerda,B.,Pirzkal,N.,Burgasser,A.,&Hsu,C.-C.2023,arXiv e-prints,arXiv:2306.12363 Itoh,Y.,Tamura,M.,&Nakajima,T.1999,AJ,117,1471 Kang,G.,Lim,J.,&Seo,B.2024,arXive-prints,arXiv:2409.04490 Karpov,S.,Malkov,O.,&Avdeeva,A.2025,A&A,695,A195 Kirkpatrick,J...

work page arXiv 2021

[13] [13]

F., Marsh, K

Skrutskie, M. F., Marsh, K. A., Wright, E. L., Mainzer, A., Eisenhardt,P.R.,McLean,I.S.,Thompson,M.A.,Bauer,J.M., Benford,D.J.,Bridge,C.R.,Lake,S.E.,Petty,S.M.,Stanford, S. A., Tsai, C.-W., Bailey, V., Beichman, C. A., Bloom, J. S., 40 Callen et al. Spectral classification of brown dwarfs using machine learning Bochanski,J.J.,Burgasser,A.J.,Capak,P.L.,Cru...

work page 2011

[14] [14]

Gonzales, E. C., Hsu, C.-C., Kamraj, N., Kiman, R., Rees, J., Theissen,C.,Ammar,K.,Andersen,N.S.,Beaulieu,P.,Colin, G.,Elachi,C.A.,Goodman,S.J.,Gramaize,L.,Hamlet,L.K., Hong,J.,Jonkeren,A.,Khalil,M.,Martin,D.W.,Pendrill,W., Pumphrey,B.,Rothermich,A.,Sainio,A.,Stenner,A.,Tanner, C., Thévenot, M., Voloshin, N. V., Walla, J., Wędracki, Z., & BackyardWorlds: ...

work page 2021

[15] [15]

P., Smith, J

Schneider, D. P., Smith, J. A., Strauss, M. A., Uomoto, A., & York,D.G.2002,ApJ,564,452 Leggett,S.K.,Tremblin,P.,Esplin,T.L.,Luhman,K.L.,&Morley, C.V.2017,ApJ,842,118 Marley,M.S.,Saumon,D.,Visscher,C.,Lupu,R.,Freedman,R., Morley,C.,Fortney,J.J.,Seay,C.,Smith,A.J.R.W.,Teal,D.J., &Wang,R.2021,ApJ,920,85 Meisner,A.M.,Caselden,D.,Schlafly,E.F.,&Kiwy,F.2023,AJ...

work page 2002

[16] [16]

Monin, J. L., Guieu, S., Pinte, C., Rebull, L., Goldsmith, P., Fukagawa,M.,Ménard,F.,Padgett,D.,Stappelfeld,K.,McCabe, C.,Carey,S.,Noriega-Crespo,A.,Brooke,T.,Huard,T.,Terebey, S.,Hillenbrand,L.,&Guedel,M.2010,A&A,515,A91 Nakajima,T.,Oppenheimer,B.R.,Kulkarni,S.R.,Golimowski, D.A.,Matthews,K.,&Durrance,S.T.1995,Nature,378,463 Pedregosa,F.,Varoquaux,G.,Gra...

work page 2010

[17] [17]

M., Drummond, B., & Hébrard, E

Spiegelman, F., Goyal, J. M., Drummond, B., & Hébrard, E. 2020,A&A,637,A38 Plewa,P.M.2018,MNRAS,476,3974 Rilinger,A.M.&Espaillat,C.C.2021,ApJ,921,182 Saumon,D.&Marley,M.S.2008,ApJ,689,1327 Schneider,A.C.,Munn,J.A.,Vrba,F.J.,Bruursema,J.,Dahm, S.E.,Williams,S.J.,Liu,M.C.,&Dorland,B.N.2023,AJ,166, 103

work page 2020

[18] [18]

M., Beichman, C., Capps, R., Chester,T.,Elias,J.,Huchra,J.,Liebert,J.,Lonsdale,C.,Monet, D.G.,Price,S.,Seitzer,P.,Jarrett,T.,Kirkpatrick,J.D.,Gizis, J

Schneider, S., Carpenter, J. M., Beichman, C., Capps, R., Chester,T.,Elias,J.,Huchra,J.,Liebert,J.,Lonsdale,C.,Monet, D.G.,Price,S.,Seitzer,P.,Jarrett,T.,Kirkpatrick,J.D.,Gizis, J. E., Howard, E., Evans, T., Fowler, J., Fullmer, L., Hurt, R., Light,R.,Kopan,E.L.,Marsh,K.A.,McCallon,H.L.,Tam,R., VanDyk,S.,&Wheelock,S.2006,AJ,131,1163

work page 2006

[19] [19]

J., Faherty, J

Skrzypek, N., Warren, S. J., Faherty, J. K., Mortlock, D. J., Burgasser,A.J.,&Hewett,P.C.2015,A&A,574,A78 Suárez,G.&Metchev,S.2022,MNRAS,513,5701 Suárez,G.,Vos,J.M.,Metchev,S.,Faherty,J.K.,&Cruz,K.2023, ApJ,954,L6 Tarter,J.C.1975,PhDthesis,UniversityofCalifornia,Berkeley Testi,L.,Natta,A.,Scholz,A.,Tazzari,M.,Ricci,L.,&deGregorio Monsalvo,I.2016,A&A,593,A...

work page 2015

[20] [20]

J., Blain, A., Mendez, B., Irace, W

McLean, I., Benford, D., Lonsdale, C. J., Blain, A., Mendez, B., Irace, W. R., Duval, V., Liu, F., Royer, D., Heinrichsen, I., Howard,J.,Shannon,M.,Kendall,M.,Walsh,A.L.,Larsen,M., Cardon,J.G.,Schick,S.,Schwalm,M.,Abid,M.,Fabinsky,B., Naes,L.,&Tsai,C.-W.2010,AJ,140,1868 Ye,G.,Zhang,H.,&Wu,Q.2024,ApJS,275,19 APPENDICES A. Appendix: Spectral types predicted...

work page 2010