arxiv: 2603.06011 · v2 · submitted 2026-03-06 · ❄️ cond-mat.mtrl-sci

Recognition: 2 theorem links

· Lean Theorem

Spectra-Scope : A toolkit for automated and interpretable characterization of material properties from spectral data

Amalya C. Johnson , Chris Fajardo , Leena Sansguiri , Weike Ye , Steven B. Torrisi

Authors on Pith no claims yet

Pith reviewed 2026-05-15 15:33 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci

keywords spectroscopymachine learningAutoMLinterpretabilitymaterial characterizationspectral datafeature extractionnonlinear transformations

0 comments

The pith

Spectra-Scope automates interpretable machine learning models from spectral data to characterize material properties.

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

Spectroscopy yields information on material structure, composition, and dynamics, yet nonlinear correlations between signals and target properties often hinder the creation of reliable supervised models. Spectra-Scope supplies an open-source AutoML framework that performs data preprocessing, nonlinear feature extraction, model training, and automated feature downselection to produce simple, interpretable models. The toolkit runs in Python or through a no-code web application and requires only modest computational resources. It matches the predictive performance of comparable literature models on materials and agricultural spectroscopy datasets. Its emphasis on interpretability lets users explain individual model outputs and link spectral features to underlying physical processes.

Core claim

Spectra-Scope is an AutoML toolkit that applies nonlinear feature transformations followed by automated downselection to train simple, interpretable machine learning models on spectroscopy data. This pipeline reproduces the accuracy of comparable models in the literature while enabling rationalization of model behavior and direct connection of selected features to physical processes in materials and agricultural samples.

What carries the argument

Nonlinear feature extraction combined with automated feature downselection inside an interpretable ML training loop that supports rapid model development with modest resources.

If this is right

Users can train multiple interpretable models on a set of feature transformations quickly without specialized hardware.
Interpretability tools allow rationalization of why individual models succeed or fail on particular spectra.
Physical origins of spectral features become accessible through the downselected model explanations.
The same workflow applies equally to materials characterization and agricultural spectroscopy tasks.

Where Pith is reading between the lines

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

Domain scientists without deep machine-learning expertise could apply the toolkit directly to new spectral instruments.
The interpretability focus may reduce reliance on black-box models in experimental workflows where physical insight is required.
Integration with streaming or in-situ spectroscopy setups could enable near-real-time property tracking.
The nonlinear transformation library might be reused as a modular component in other AutoML pipelines for scientific data.

Load-bearing premise

The chosen nonlinear feature transformations and automated downselection will reliably produce models that stay accurate and genuinely interpretable across varied spectral datasets without discarding essential physical information or creating spurious correlations.

What would settle it

A spectral dataset on which Spectra-Scope reaches literature-comparable accuracy yet the downselected features show no correspondence to known physical spectral signatures or produce interpretability explanations that contradict established material physics.

Figures

Figures reproduced from arXiv: 2603.06011 by Amalya C. Johnson, Chris Fajardo, Leena Sansguiri, Steven B. Torrisi, Weike Ye.

**Figure 1.** Figure 1: Outline of this paper and the Spectra-Scope pipeline. (a) Input data can come from any experimental or simulated 1-D array data source for inference on a scalar response variable. (b) Available featurizations of spectral data include the cumulative distribution function, Gaussian peak fitting, principal component analysis, polynomial peak fitting, and others as outlined in the methods. (c) Transformed spec… view at source ↗

**Figure 2.** Figure 2: Front page of Spectra-Scope application. Multiple data types can be input and visualized on the home page. The app includes abilities to featurize data, visualize featurizations, train models using random forests or LCEN, and visualize the important or downselected features by the model. Application Spectra-Scope is available as a no-code web-based application at spectrascope.matr.io (See [PITH_FULL_IMAGE… view at source ↗

**Figure 3.** Figure 3: Regressing mean nearest-neighbor distance from simulated XANES spectra and PDFs of Ti-oxide structures. (a) Summary of RMSE for regressing bond length using LCEN and random forests for XANES, PDF, XANES + PDF, and other transformations of the data. CDF: cumulative distribution function. NLTrans: Nonlinear feature expansion as outlined in the main text and the supplementary information. Clustering: Feature … view at source ↗

**Figure 4.** Figure 4: Regressing grape sugar content. (a) % RMSE for random forests and LCEN models built on Vis-NIR and Raman spectra transformed in different ways. (b) Top 10 most important features for predicting TSS with the full spectrum (i) and polynomial features extracted from the spectrum (ii) using random forests. (c) 10 highest absolute magnitude coefficients for regressing TSS with the full spectrum (i) and polynomi… view at source ↗

**Figure 5.** Figure 5: Fused LASSO selected Features. The top panel shows all of the NIR spectra in the dataset. The bottom panel shows the regression coefficients for fused LASSO models with different regularization parameters α. 10/15 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗

**Figure 6.** Figure 6: Linear Correlation Assessment. Analysis for (a) XANES + PDF and (b) Grapes dataset for predicting their respective target variable. r : Pearson’s correlation coefficient. Right: Corresponding metric for each model. 11/15 [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗

read the original abstract

Spectroscopy is a central pillar of materials characterization, providing useful information on properties like structure, composition, or excited state dynamics of a system. However, many spectroscopic techniques present challenges in development of interpretable, performant, and reliable supervised learning models due to the wide range of possible nonlinear correlations that can exist between the signal and the response variable (target) of interest. Here, we present Spectra-Scope, an open-source AutoML framework for automatic characterization of material properties from spectroscopy data using interpretable machine learning (ML) models. The software is implemented in Python and a no-code web application. It comprises tools for data preprocessing, nonlinear feature extraction, machine learning model training, and feature downselection. Users can easily train different types of simple, interpretable ML models on a set of feature transformations quickly and with modest computational resources. In this work, we outline the methods of Spectra-Scope and its effectiveness across diverse datasets, with applications to materials and agricultural spectroscopy data. We show that Spectra-Scope can reproduce performance of comparable models in the literature, and highlight how our emphasis on interpretability can be used to rationalize the behavior of individual models and understand the physical processes behind spectral features.

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

Spectra-Scope bundles existing AutoML steps into a usable Python toolkit with a web app for spectral data, but the abstract gives no metrics to back its performance or interpretability claims.

read the letter

Spectra-Scope is a Python package plus no-code web app that strings together preprocessing, nonlinear feature transforms such as polynomials or kernels, training of simple interpretable models, and automated feature downselection for spectroscopy data in materials and agricultural contexts. The authors position it as a practical way to handle the messy correlations common in spectral signals without heavy custom coding. That integration is the main new piece: a single named, open-source pipeline aimed at experimental users rather than a novel algorithm. The focus on keeping models interpretable so users can map features back to physical processes is a reasonable practical choice and could help routine analysis. The paper claims it reproduces literature performance on diverse datasets, which would be useful if true. The soft spots sit in the missing evidence. The abstract asserts reproduction of prior results and useful interpretability but supplies no numbers, dataset names, validation splits, or ablation checks. Without those, it is impossible to judge whether the nonlinear expansions preserve real spectroscopic meaning or whether the downselection avoids dataset-specific noise. The stress-test worry about spurious correlations is fair to raise until the full results show explicit comparisons to known peak positions or physical models. This paper is for lab researchers who want a ready pipeline for standard spectral property prediction rather than theorists or method developers. A reader who needs code and a demo would get some value from the implementation. It deserves peer review so the benchmarks and code can be examined directly.

Referee Report

2 major / 2 minor

Summary. The manuscript presents Spectra-Scope, an open-source Python AutoML toolkit (with accompanying no-code web application) for spectral data preprocessing, nonlinear feature extraction (e.g., polynomials and kernels), training of simple interpretable ML models, and automated feature downselection. It applies the framework to materials and agricultural spectroscopy datasets and claims that the toolkit reproduces the performance of comparable literature models while enabling interpretability to rationalize individual model behavior and link selected features to underlying physical processes.

Significance. If the performance and interpretability claims are substantiated with quantitative evidence, Spectra-Scope would offer a practical, accessible tool that lowers the barrier for materials scientists to generate accurate yet physically grounded models from spectroscopic data. The emphasis on interpretability addresses a recognized gap in black-box spectral ML applications and could facilitate discovery of structure-property relationships.

major comments (2)

[Results] Results section: the central claim that Spectra-Scope 'reproduces performance of comparable models in the literature' is unsupported by any quantitative metrics, specific datasets, validation protocols, ablation studies, or direct numerical comparisons (e.g., no reported RMSE, R², or accuracy values against published baselines).
[Methods] Methods (nonlinear feature extraction and downselection): the automated transformations and statistical downselection are presented without systematic validation against known spectroscopic assignments (peak positions, band structures, or physical models); this leaves open the risk that selected features reflect dataset-specific correlations rather than causal physical processes, directly undermining the interpretability benefit asserted in the abstract.

minor comments (2)

[Abstract] Abstract and introduction: the phrase 'diverse datasets' is used without naming the specific materials or agricultural spectra employed or citing the corresponding literature benchmarks.
[Figures] Figure captions and text: notation for feature transformations (e.g., polynomial order, kernel parameters) is introduced without a concise summary table, making it difficult to reproduce the exact pipeline.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and valuable comments, which have helped us improve the manuscript. We address each major comment below and outline the revisions we plan to make.

read point-by-point responses

Referee: [Results] Results section: the central claim that Spectra-Scope 'reproduces performance of comparable models in the literature' is unsupported by any quantitative metrics, specific datasets, validation protocols, ablation studies, or direct numerical comparisons (e.g., no reported RMSE, R², or accuracy values against published baselines).

Authors: We acknowledge that the Results section does not provide direct quantitative comparisons to literature models, which is necessary to fully support the claim. The manuscript does present performance metrics on the datasets used, but without explicit benchmarking. In the revised manuscript, we will include quantitative comparisons, specifying the datasets, reporting RMSE, R², and accuracy values, detailing the validation protocols (e.g., cross-validation), and including ablation studies where appropriate. revision: yes
Referee: [Methods] Methods (nonlinear feature extraction and downselection): the automated transformations and statistical downselection are presented without systematic validation against known spectroscopic assignments (peak positions, band structures, or physical models); this leaves open the risk that selected features reflect dataset-specific correlations rather than causal physical processes, directly undermining the interpretability benefit asserted in the abstract.

Authors: This is a valid concern. The feature extraction and downselection in Spectra-Scope are designed to be automated and data-driven to prioritize predictive performance while maintaining interpretability through simple models. However, we agree that without linking to physical assignments, the interpretability claims are weakened. We will revise the Methods section to describe how users can validate features against known assignments and add examples in the Results where selected features correspond to established spectroscopic features in the materials and agricultural datasets. revision: yes

Circularity Check

0 steps flagged

Software toolkit demonstration with no derivation chain

full rationale

The paper describes an open-source AutoML framework (Spectra-Scope) for preprocessing, nonlinear feature extraction, model training, and downselection on spectral data. It reports empirical performance on materials and agricultural datasets and claims reproducibility of literature results plus interpretability benefits. No equations, predictions, or first-principles derivations are presented that could reduce to fitted inputs by construction. No self-citations are invoked as uniqueness theorems, and no ansatzes or renamings of known results are used to support central claims. The work is implementation and demonstration rather than a closed mathematical argument, so no circular steps exist.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The contribution rests on standard supervised learning assumptions rather than new physical axioms or invented entities.

axioms (1)

domain assumption Nonlinear feature transformations of spectral data can be combined with simple interpretable models to capture relevant material-property relationships
The framework presupposes that such transformations preserve enough physical signal to allow both accurate prediction and human-interpretable feature importance.

pith-pipeline@v0.9.0 · 5533 in / 1251 out tokens · 63829 ms · 2026-05-15T15:33:19.027333+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages · 2 internal anchors

[1]

J.et al.An autonomous laboratory for the accelerated synthesis of novel materials.Nature624, 86–91, DOI: 10.1038/s41586-023-06734-w (2023)

Szymanski, N. J.et al.An autonomous laboratory for the accelerated synthesis of novel materials.Nature624, 86–91, DOI: 10.1038/s41586-023-06734-w (2023)

work page doi:10.1038/s41586-023-06734-w 2023
[2]

M., Zhou, L

Gregoire, J. M., Zhou, L. & Haber, J. A. Combinatorial synthesis for AI-driven materials discovery.Nat. Synth.2, 493–504, DOI: 10.1038/s44160-023-00251-4 (2023)

work page doi:10.1038/s44160-023-00251-4 2023
[3]

Hello World

Baird, S. G. & Sparks, T. D. Building a “Hello World” for self-driving labs: The Closed-loop Spectroscopy Lab Light-mixing demo.STAR Protoc.4, 102329, DOI: 10.1016/j.xpro.2023.102329 (2023)

work page doi:10.1016/j.xpro.2023.102329 2023
[4]

d., Ehsani, R., Brillante, L

Ebrahimi, I., Castro, R. d., Ehsani, R., Brillante, L. & Feng, S. Advancing grape chemical analysis through machine learning and multi-sensor spectroscopy.J. Agric. F ood Res.16, 101085, DOI: 10.1016/j.jafr.2024.101085 (2024). 12/15

work page doi:10.1016/j.jafr.2024.101085 2024
[5]

& Wang, Z

Tan, C., Wu, H., Yang, L. & Wang, Z. Cutting edge high-throughput synthesis and characterization techniques in combinatorial materials science.Adv. Mater. Technol.9, 2302038, DOI: 10.1002/admt.202302038 (2024)

work page doi:10.1002/admt.202302038 2024
[6]

Takeuchi, I.et al.Data management and visualization of x-ray diffraction spectra from thin film ternary composition spreads.Rev. Sci. Instruments76, 062223, DOI: 10.1063/1.1927079 (2005)

work page doi:10.1063/1.1927079 2005
[7]

Ogunlade, B.et al.Rapid, antibiotic incubation-free determination of tuberculosis drug resistance using machine learning and raman spectroscopy.Proc. Natl. Acad. Sci.121, e2315670121, DOI: 10.1073/pnas.2315670121 (2024)

work page doi:10.1073/pnas.2315670121 2024
[8]

& Ago, H

Solís-Fernández, P. & Ago, H. Machine learning determination of the twist angle of bilayer graphene by raman spectroscopy: Implications for van der waals heterostructures.ACS Appl. Nano Mater.5, 1356–1366, DOI: 10.1021/acsanm.1c03928 (2022)

work page doi:10.1021/acsanm.1c03928 2022
[9]

& Meunier, V

Sheremetyeva, N., Lamparski, M., Daniels, C., Van Troeye, B. & Meunier, V . Machine-learning models for raman spectra analysis of twisted bilayer graphene.Carbon169, 455–464, DOI: 10.1016/j.carbon.2020.06.077 (2020)

work page doi:10.1016/j.carbon.2020.06.077 2020
[10]

Adv.11, eadu7426, DOI: 10.1126/sciadv.adu7426 (2025)

Liang, H.et al.Real-time experiment-theory closed-loop interaction for autonomous materials science.Sci. Adv.11, eadu7426, DOI: 10.1126/sciadv.adu7426 (2025)

work page doi:10.1126/sciadv.adu7426 2025
[11]

Commun.11, 1706, DOI: 10.1038/s41467-020-15235-7 (2020)

Zhang, Y .et al.Identifying degradation patterns of lithium ion batteries from impedance spectroscopy using machine learning.Nat. Commun.11, 1706, DOI: 10.1038/s41467-020-15235-7 (2020)

work page doi:10.1038/s41467-020-15235-7 2020
[12]

J., K, R

Joy, N. J., K, R. M. & Balakrishnan, J. A simple and robust machine learning assisted process flow for the layer number identification of TMDs using optical contrast spectroscopy.J. Physics: Condens. Matter35, 025901, DOI: 10.1088/1361-648X/ac9f96 (2022)

work page doi:10.1088/1361-648x/ac9f96 2022
[13]

Tsamardinos, I.et al.An Automated Machine Learning architecture for the accelerated prediction of Metal-Organic Frameworks performance in energy and environmental applications.Microporous Mesoporous Mater.300, 110160, DOI: 10.1016/j.micromeso.2020.110160 (2020)

work page doi:10.1016/j.micromeso.2020.110160 2020
[14]

Ji, Z.et al.Research and application validation of a feature wavelength selection method based on acousto-optic tunable filter (aotf) and automatic machine learning (automl).Materials15, DOI: 10.3390/ma15082826 (2022)

work page doi:10.3390/ma15082826 2022
[15]

& Braatz, R

Sun, W. & Braatz, R. D. Smart process analytics for predictive modeling.Comput. & Chem. Eng.144, 107134, DOI: 10.1016/j.compchemeng.2020.107134 (2021)

work page doi:10.1016/j.compchemeng.2020.107134 2020
[16]

& Braatz, R

Sun, W. & Braatz, R. D. ALVEN: Algebraic learning via elastic net for static and dynamic nonlinear model identification. Comput. & Chem. Eng.143, 107103, DOI: 10.1016/j.compchemeng.2020.107103 (2020)

work page doi:10.1016/j.compchemeng.2020.107103 2020
[17]

& Kempa-Liehr, A

Christ, M., Braun, N., Neuffer, J. & Kempa-Liehr, A. W. Time series feature extraction on basis of scalable hypothesis tests (tsfresh – a python package).Neurocomputing307, 72–77, DOI: 10.1016/j.neucom.2018.03.067 (2018). 18.Breiman, L. Random Forests.Mach. Learn.45, 5–32, DOI: 10.1023/A:1010933404324 (2001)

work page doi:10.1016/j.neucom.2018.03.067 2018
[18]

Chen, Y .et al.Robust Machine Learning Inference from X-ray Absorption Near Edge Spectra through Featurization. Chem. Mater.36, 2304–2313, DOI: 10.1021/acs.chemmater.3c02584 (2024)

work page doi:10.1021/acs.chemmater.3c02584 2024
[19]

B.et al.Random forest machine learning models for interpretable X-ray absorption near-edge structure spectrum-property relationships.npj Comput

Torrisi, S. B.et al.Random forest machine learning models for interpretable X-ray absorption near-edge structure spectrum-property relationships.npj Comput. Mater.6, 109, DOI: 10.1038/s41524-020-00376-6 (2020)

work page doi:10.1038/s41524-020-00376-6 2020
[20]

& Tsitsiklis, J.Introduction to Probability

Bertsekas, D. & Tsitsiklis, J.Introduction to Probability. Athena Scientific optimization and computation series (Athena Scientific, 2008)

work page 2008
[21]

On measures of dependence.Acta Math

Rényi, A. On measures of dependence.Acta Math. Acad. Sci. Hungarica10, 441–451, DOI: 10.1007/BF02024507 (1959)

work page doi:10.1007/bf02024507 1959
[22]

& Celisse, A

Arlot, S. & Celisse, A. A survey of cross validation procedures for model selection.Stat. Surv.4, DOI: 10.1214/09-SS054 (2009)

work page doi:10.1214/09-ss054 2009
[23]

A Unified Approach to Interpreting Model Predictions

Lundberg, S. & Lee, S.-I. A unified approach to interpreting model predictions. Preprint at arXiv (2017). https: //arxiv.org/abs/1705.07874

work page internal anchor Pith review Pith/arXiv arXiv 2017
[24]

Consistent Individualized Feature Attribution for Tree Ensembles

Lundberg, S. M., Erion, G. G. & Lee, S.-I. Consistent individualized feature attribution for tree ensembles. Preprint at arXiv (2019). https://arxiv.org/abs/1802.03888. 26.Pedregosa, F.et al.Scikit-learn: Machine learning in Python.J. Mach. Learn. Res.12, 2825–2830 (2011)

work page internal anchor Pith review Pith/arXiv arXiv 2019
[25]

& Braatz, R

Seber, P. & Braatz, R. D. LCEN: A novel feature selection algorithm for nonlinear, interpretable machine learning models. Preprint at arXiv (2024). https://arxiv.org/abs/2402.17120

work page arXiv 2024
[26]

Regression shrinkage and selection via the lasso.J

Tibshirani, R. Regression shrinkage and selection via the lasso.J. Royal Stat. Soc. Ser. B (Methodological)58, 267–288 (1996). 13/15

work page 1996
[27]

Tibshirani, R. J. & Taylor, J. The solution path of the generalized lasso.The Annals Stat.39, DOI: 10.1214/11-AOS878 (2011)

work page doi:10.1214/11-aos878 2011
[28]

Rhyu, J.et al.Systematic feature design for cycle life prediction of lithium-ion batteries during formation.Joule9, 101884, DOI: https://doi.org/10.1016/j.joule.2025.101884 (2025)

work page doi:10.1016/j.joule.2025.101884 2025
[29]

APL materials1(2013)

Jain, A.et al.Commentary: The materials project: A materials genome approach to accelerating materials innovation. APL materials1(2013)

work page 2013
[30]

data5, 1–8, DOI: 10.1038/sdata

Mathew, K.et al.High-throughput computational x-ray absorption spectroscopy.Sci. data5, 1–8, DOI: 10.1038/sdata. 2018.151 (2018)

work page doi:10.1038/sdata 2018
[31]

Mater.4, DOI: 10.1038/s41524-018-0067-x (2018)

Zheng, C.et al.Automated generation and ensemble-learned matching of X-ray absorption spectra.npj Comput. Mater.4, DOI: 10.1038/s41524-018-0067-x (2018)

work page doi:10.1038/s41524-018-0067-x 2018
[32]

N., Torrisi, S

Na Narong, T., Zachko, Z. N., Torrisi, S. B. & Billinge, S. J. L. Interpretable multimodal machine learning analysis of X-ray absorption near-edge spectra and pair distribution functions.npj Comput. Mater.11, 98, DOI: 10.1038/s41524-025-01589-3 (2025)

work page doi:10.1038/s41524-025-01589-3 2025
[33]

& Billinge, S

Takeshi, E. & Billinge, S. J. Chapter 3 - the method of total scattering and atomic pair distribution function analysis. In Egami, T. & Billinge, S. J. (eds.)Underneath the Bragg Peaks, vol. 16 ofPergamon Materials Series, 55–111, DOI: 10.1016/B978-0-08-097133-9.00003-4 (Pergamon, 2012)

work page doi:10.1016/b978-0-08-097133-9.00003-4 2012
[34]

T., Bunker, G., D’Angelo, P

Chantler, C. T., Bunker, G., D’Angelo, P. & Diaz-Moreno, S. X-ray absorption spectroscopy.Nat. Rev. Methods Primers4, 89, DOI: 10.1038/s43586-024-00366-8 (2024)

work page doi:10.1038/s43586-024-00366-8 2024
[35]

Moro, V .et al.Multimodal foundation models for material property prediction and discovery.Newton1, 100016, DOI: 10.1016/j.newton.2025.100016 (2025)

work page doi:10.1016/j.newton.2025.100016 2025
[36]

R.et al.Transcriptomic analysis of the late stages of grapevine (Vitis vinifera cv

Cramer, G. R.et al.Transcriptomic analysis of the late stages of grapevine (Vitis vinifera cv. Cabernet Sauvignon) berry ripening reveals significant induction of ethylene signaling and flavor pathways in the skin.BMC Plant Biol.14, 370, DOI: 10.1186/s12870-014-0370-8 (2014)

work page doi:10.1186/s12870-014-0370-8 2014
[37]

Jha, S. N. & Matsuoka, T. Non-destructive determination of acid–brix ratio of tomato juice using near infrared spectroscopy. Int. J. F ood Sci. Technol.39, 425–430, DOI: 10.1111/j.1365-2621.2004.00800.x (2004)

work page doi:10.1111/j.1365-2621.2004.00800.x 2004
[38]

Physics: Conf

Liu, Y .et al.Potable nir spectroscopy predicting soluble solids content of pears based on leds.J. Physics: Conf. Ser.277, 012026, DOI: 10.1088/1742-6596/277/1/012026 (2011)

work page doi:10.1088/1742-6596/277/1/012026 2011
[39]

& Ying, Y

Lin, H. & Ying, Y . Theory and application of near infrared spectroscopy in assessment of fruit quality: a review.Sens. Instrumentation for F ood Qual. Saf.3, 130–141, DOI: 10.1007/s11694-009-9079-z (2009)

work page doi:10.1007/s11694-009-9079-z 2009
[40]

& Lawson, P

Golic, M., Walsh, K. & Lawson, P. Short-Wavelength Near-Infrared Spectra of Sucrose, Glucose, and Fructose with Respect to Sugar Concentration and Temperature.Appl. Spectrosc.57, 139–145, DOI: 10.1366/000370203321535033 (2003)

work page doi:10.1366/000370203321535033 2003
[41]

Subramanian, J., Hung, L., Schweigert, D., Suram, S. & Ye, W. Xxact-nn: Structure agnostic multimodal learning for materials science. Preprint at arXiv (2025). https://arxiv.org/abs/2507.01054. Acknowledgements The authors thank Pedro Seber for advice on the implementation of LCEN, Richard Braatz, Tina Na Narong, and Simon Billinge for helpful discussions...

work page arXiv 2025