Pith. sign in

REVIEW 4 major objections 5 minor 27 references

Bridging Structure and Activity in Nanocatalysts via Machine Learning and Global Structure Representations

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Pith's one-line read A machine-learning workflow predicts which of 52,318 platinum nanoparticle structures are most active for oxygen reduction, using an experimentally measurable distance histogram as the input.

desk verdict PDDF-to-activity mapping is a genuine contribution, but the screening claim needs random and size-only baselines before it stands up. read the letter →

arxiv 2509.10985 v1 pith:TWJ464Y3 submitted 2025-09-13 cond-mat.mes-hall cond-mat.mtrl-sciphysics.chem-ph

classification cond-mat.mes-hallcond-mat.mtrl-sciphysics.chem-ph
keywords nanocatalysismachinelearningGaussianprocessregressionoxygenreductionreactionplatinumnanoparticlespairdistributionfunctiongeneralizedcoordinationnumberBayesianoptimization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper aims to show that a single machine-learning workflow can map nanoparticle structure directly to catalytic mass activity, using either a surface-site descriptor available from computation or a distance histogram available from scattering experiments. On a pool of 52,318 platinum nanoparticles (2–6 nm) evaluated with a microkinetic model for the oxygen reduction reaction, a Gaussian process trained on the distance-based descriptor reaches R² of about 0.67 with roughly four thousand active-learning-selected samples, while the coordination-number descriptor reaches near-perfect accuracy. With Bayesian optimization the same model recovers more than 80% of the top-10 and top-100 most active structures after probing only about three thousand candidates. If correct, this gives experimentalists a concrete route to rank and shortlist nanocatalysts from a measurable structural fingerprint rather than from atomistic simulation alone.

What carries the argument

The central objects are two global structural descriptors: the generalized coordination number (GCN) distribution, a histogram of weighted neighbor counts of surface sites that serves as a computed input to the microkinetic activity model, and the pair distance distribution function (PDDF), a histogram of interatomic distances up to twice the bulk lattice parameter that can be measured by total scattering. The learning machinery is Gaussian process regression, which provides both predictions and uncertainties, combined with active learning or Bayesian optimization using an upper confidence bound acquisition function to choose which structures to label next. The PDDF is doing the load-bearing

What would settle it

Take a random subset of the 52,318 nanoparticles, calculate or measure their PDDFs, select a top-100 shortlist with the paper's Bayesian optimization loop, measure the actual oxygen-reduction mass activity of those particles and of a randomly chosen control set under identical conditions, and compare the distributions; if the PDDF-selected set is not better than random, the central screening claim fails.

Watch

Extended reading notes

Core claim

In the authors' terms, accurate structure–activity models can be built from either the generalized coordination number distribution of the surface, which is accessible computationally, or the pair distance distribution function, which is accessible experimentally. The GCN-based model is near-perfect (R² = 0.959) because the ground-truth microkinetic model is itself a function of GCN, while the PDDF-based model reaches a useful but lower accuracy (R² = 0.672). On top of prediction, Bayesian optimization guided by the PDDF-trained Gaussian process identifies over 80% of the true top-10 and top-100 activities out of 52,318 candidates after roughly 300 acquisition steps, with only small highly a

Load-bearing premise

The rankings are only as trustworthy as the microkinetic model that produced the activity labels, and that model's ordering of real catalysts has not been checked experimentally here.

Editorial extensions

If this is right

  • An experimentally measurable quantity, the PDDF, can be used quantitatively, not just qualitatively, to rank nanoparticle catalysts by mass activity.
  • A training budget of a few thousand carefully chosen samples is enough to reduce a 52,318-candidate search to a shortlist containing most of the best structures.
  • Because the same Gaussian-process plus Bayesian-optimization loop works with either descriptor, computational prescreening and experimental characterization can feed the same model.
  • The workflow is expected to extend to other activity models, such as kinetic Monte Carlo, and to other reactions, multi-element materials, and supported architectures.
  • The residual misclassification of around 300-atom particles indicates that small-size effects remain an open issue for PDDF-based screening.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's success rate suggests a practical experimental protocol: measure PDDFs for a few thousand candidates, label a subset by activity, and let the same loop pick the next measurements; this would test whether the ranking survives real electrochemical conditions.
  • If PDDF-ranking holds experimentally, the same approach could transfer to alloy and supported nanoparticles, where scattering already encodes strain and ordering; the paper does not claim this, but its descriptors are not platinum-specific.
  • The GCN result sets an upper bound: since the ground-truth labels are generated from GCN, no representation can do better than GCN in this closed loop, so the PDDF's gap of about 0.29 in R² measures the information lost when only an experimentally accessible fingerprint is used.
  • A natural next step would be a size-dependent correction or a hybrid descriptor for nanoparticles below roughly 300 atoms, where the Bayesian optimization systematically misses the most active structures.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a Gaussian-process-regression (GPR) workflow that maps global structural descriptors of Pt nanoparticles to their computed mass activity for the oxygen reduction reaction, and uses the resulting surrogate in a Bayesian optimization (BO) loop to screen a fixed pool of 52,318 nanoparticle structures. Two descriptors are compared: the distribution of generalized coordination numbers (GCN) of surface sites, which is the same descriptor used as input to the microkinetic model that generated the ground-truth labels, and the pair distance distribution function (PDDF), which is experimentally accessible. With active learning, the GCN-based model reaches R2=0.959 and the PDDF-based model reaches R2=0.672 at 4,020 training points. The BO exercise, using a UCB acquisition function with κ=2 and batches of 10 acquisitions, reports recovery of at least 80% of the top-10 and top-100 most active structures after roughly 3,000 additional evaluations. The authors frame the work as a blueprint for accelerated theoretical and experimental nanocatalyst discovery.

Significance. If the claims hold, the most valuable contribution is the demonstration that a measurable distance-based representation (PDDF) can support predictive screening of nanoparticle catalysts, with the GCN model serving as an upper-bound sanity check. The paper is honest about several limitations: it explicitly acknowledges that the GCN result is expected because the ground-truth labels were generated from GCN distributions, it reports that the PDDF model does not saturate within the training sizes considered, and it notes that misclassifications occur for the smallest highly active particles. The availability of code, data, and reproducibility scripts is a clear strength. However, the central screening claim—'efficiently identifies the Top-10 and Top-100'—is not yet supported without comparison to simple baselines; the present evidence is compatible with the possibility that much of the apparent efficiency is due to a trivial particle-size signal.

major comments (4)
  1. [Figure 4 and 'Bayesian optimization' section (pp. 8-9)] The BO claim needs baselines. The text reports that after 300 acquisition steps (about 3,000 additional evaluations) at least 80% of the top-10 and top-100 are recovered, but no comparison is made to (i) random acquisition with the same budget, or (ii) a model using only particle size. Because the target is mass activity, smaller particles are strongly favored by surface-area-to-mass ratio; the top-10 shown in Figure 4 are small, and the text itself notes that misclassifications occur only for particles near 300 atoms. PDDF histograms encode size through peak positions and amplitudes, so a size-only surrogate could plausibly recover a large fraction of the top-100 without learning any structure–activity relationship beyond size. I request concrete controls: report top-k recall as a function of acquisitions for random selection and for a GPR/linear model trained on number of atoms (or ano
  2. [SI Figure S12 and 'Bayesian optimization' section (κ=2)] The UCB exploration weight κ=2 is 'empirically observed to strike the best balance', with the comparison shown in SI Figure S12. If this tuning is performed on the same candidate pool whose top-100 recovery is then reported, the reported trace is optimistically selected. The manuscript should either report the BO performance for several κ values on a held-out subset of the pool, or show that the qualitative conclusion (PDDF-driven BO substantially outperforms random/size baselines) is insensitive to κ. The current presentation does not allow the reader to separate algorithmic merit from tuning luck.
  3. [Abstract and Figure 3 (PDDF accuracy)] The abstract says 'Accurate models are obtained when leveraging either ... GCN ... or ... PDDF', but the PDDF model reaches only R2=0.672 with no saturation in the learning curve (Figure 3, middle panel) and MAE=0.467 A/mg. The GCN model's high R2 is, by the paper's own admission in the conclusion, an expected consequence of the label-generation process. The nontrivial predictive claim therefore rests entirely on the PDDF model, and its moderate R2 should not be described as 'accurate' without additional evidence that this level of error is sufficient for the intended ranking task. I recommend reporting rank-based metrics (e.g., Spearman correlation, top-k precision) and tempering the abstract's wording unless the BO baselines in the previous comment show that the PDDF model has genuine ranking power.
  4. [Conclusions (p. 9-10)] The manuscript states that 'Accurate machine learning predictions ... were expected ... since these same were used as inputs in the microkinetic model that generated the ground truth data.' This statement should be moved earlier (or mirrored in the abstract) so that readers do not interpret the GCN R2=0.959 as evidence of an independent structure–activity mapping. The disclosure is commendable, but the current abstract's phrasing 'accurate models are obtained' for the GCN representation invites a misleading reading. This is a presentation issue, but it directly affects the paper's central claim as advertised.
minor comments (5)
  1. [Author information (p. 10)] Typo: 'Authos' should be 'Authors'.
  2. [Figure 4 caption and BO text (pp. 8-9)] The caption refers to 'four Bayesian optimization runs' in the bottom panel, but the top panel appears to report percentages per iteration without specifying whether they are means, medians, or single representative runs. Please clarify and, if available, include standard deviations or confidence bands.
  3. [Data availability statement (p. 11)] The XYZ coordinates are listed as available 'upon acceptance at: *zenodo link*'; the placeholder should be replaced with the actual DOI before publication.
  4. [Main text, BO section (p. 8)] The sentence 'Within just 200 acquisition steps - where 10 structures from the candidate pool are selected at each acquisition step' is clear, but it would help to state explicitly that this corresponds to 2,000 additional activity evaluations plus the initial 10 training points, so the reader can compare with the abstract's 'few thousand structures'.
  5. [General] Some SI figure references (Figures S5-S12) are mentioned in the text without describing the actual content in the main text; for a journal readership, one sentence each on what those robustness checks show would improve readability.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: GCN-leg accuracy is built-in; PDDF/BO screening remains independent.

  1. self definitional [Conclusion (page 9-10)]
    "Accurate machine learning predictions of structure–activity relationships were expected when representing nanoparticles using the distribution of atop generalized coordination numbers of their surface sites, since these same were used as inputs in the microkinetic model that generated the ground truth data."

    The ground-truth mass activities are computed by a microkinetic model (ref 12) that takes the GCN distribution as input. The GPR trained on the same GCN distribution to predict those labels is fitting a surrogate to the label-generating function. The reported R²=0.959 thus measures how well the GPR reproduces the microkinetic model, not an independent structure–activity correlation. The paper admits this expectation, but the abstract still lists GCN-based accuracy as one of the two accurate representations, making the GCN leg a self-definitional demonstration. The PDDF model and BO screening do not reduce in the same way because PDDF is not the input to the microkinetic model.

full rationale

The paper's central novel claim—that an experimentally accessible PDDF can be used for quantitative activity prediction and Bayesian-optimization screening—is not circular: the PDDF is not the input to the microkinetic model that generated the labels, and the reported R²=0.672 is a non-trivial result about information content. The GCN-based model, however, is circular by construction: it predicts from the very descriptor used to compute the ground truth, as the authors explicitly acknowledge. This partial circularity is localized to the GCN proof-of-concept and does not undermine the PDDF/BO results. The use of the self-authored microkinetic model (ref 12) as ground truth is a legitimate external model, not a circular step, and the absence of random/size baselines in the BO analysis is a rigor concern rather than a circularity. Overall score reflects the built-in GCN 'prediction' while recognizing the independent content of the PDDF pathway.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

All accuracy and screening statistics are measured against a single microkinetic model (ref 12) that maps atop-site GCN distributions to MA@0.9V. The GCN-based ML leg uses the exact descriptor that generates the labels, so its R2=0.959 is a consistency check; the PDDF leg and the BO loop are the parts that genuinely test generalization to a different representation. Free parameters are histogram resolutions, UCB weight kappa, and protocol sizes, all chosen by hand against the same dataset. No new physical entities are introduced.

free parameters (5)
  • UCB exploration weight kappa = 2
    Selected empirically as the best exploration-exploitation tradeoff; kappa=10 and kappa=0.2 are reported as less efficient (SI Figure S12). Tuned on the same screening dataset, so reported BO efficiency partly depends on this choice.
  • GCN distribution bin width = 0.25
    Histogram resolution chosen by hand; the authors report that model accuracy varies with this resolution (SI Figure S5).
  • PDDF bin width = 0.01 lattice parameter
    Histogram resolution for the PDDF representation, chosen by hand; trends are reported to hold across binning choices (SI Figures S6-S11).
  • PDDF cut-off distance = 2 x bulk lattice parameter
    Truncation of the pair-distance histogram, chosen by hand; robustness explored in SI.
  • BO protocol sizes: initial samples and acquisitions per step = 10 initial, 10 per step
    Hand-chosen protocol sizes that set the x-axis of the learning curves and the 'few thousand probes' framing of the screening claim.
assumptions (5)
  • domain assumption The microkinetic model of ref 12 correctly estimates eORR mass activity at 0.9V from atop-site GCN distributions, and its outputs are valid ground-truth labels and optimization targets.
    Invoked throughout (Structure-Activity Theory and Modeling in SI; Figures 2-4). Every R2, MAE, and top-k fraction is computed against this model. External validity of the screening claim depends on it, and it is not experimentally validated here.
  • domain assumption GCN scaling relations for eORR on Pt (refs 13, 14, 17) hold for nanoparticles of 264-2830 atoms, including the activity peak assigned to GCN=8.33.
    These relations are the physics inside the ground-truth model; the workflow inherits their validity. Introduced in the Introduction and SI.
  • domain assumption The PDDF histogram at 0.01 lattice-parameter resolution with a 2x lattice cutoff preserves enough structural information to infer the GCN content relevant to activity.
    This is the key representational bet of the paper; the moderate R2=0.672 is the empirical test of it. Introduced in the PDDF section and Figure 3.
  • standard math GPR with the chosen kernel and marginal-likelihood hyperparameter fitting is an appropriate and sufficiently expressive regressor for these structure-activity maps.
    Standard GP practice (ref 15); not load-bearing for the central claim.
  • domain assumption The 52,318-structure candidate pool assembled from prior investigations is a representative design space for Pt nanoparticle eORR screening.
    Defines the 'Top-10 / Top-100 out of more than 50000' claim; pool composition is fixed by earlier studies (SI Dataset Generation).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging Structure and Activity in Nanocatalysts via Machine Learning and Global Structure Representations." pith.science (2026). https://pith.science/paper/TWJ464Y3

@misc{pith2026250910985,
  author       = {Pith},
  title        = {Pith review of: Bridging Structure and Activity in Nanocatalysts via Machine Learning and Global Structure Representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TWJ464Y3}},
  note         = {Machine review of arXiv:2509.10985}
}
read the original abstract

Establishing a mapping between nanocatalysts structure and their catalytic properties is essential for efficient design. To this end, we demonstrate the accuracy of a general machine learning framework on a representative and challenging application: predicting the mass activity of Pt nanoparticles for the electrochemical oxygen reduction reaction, estimated via a microkinetic model. Accurate models are obtained when leveraging either a nanocatalyst's structure representation accessible at the computational level, namely the surface site generalized coordination number distributions, or one accessible experimentally, namely the nanoparticle's pair distance distribution function. Building on this result, we demonstrate that our machine learning model, in tandem with Bayesian optimization, efficiently identifies the Top-10 and Top-100 most active structures out of a large pool of candidates comprising more than 50000 different structures, after probing the activity only of a few thousand structures. These findings provide a robust blueprint for accelerated theoretical and experimental identification of active nanocatalysts.

Figures

Figures reproduced from arXiv: 2509.10985 by the authors.

Figure 1
Figure 1. Graphical illustration of our workflow to bridge the modeling gap in heterogeneous [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Top) Mean GCN distribution. To estimate this quantity we average the GCN [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Top) Mean PDDF values distribution over the full dataset of structures with its [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Top) Percentage of predicted points, per iteration, that match the values of the top [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 2 linked inside Pith

  1. [1]

    M.; Xu, B

    Friend, C. M.; Xu, B. Heterogeneous catalysis: a central science for a sustainable future. Accounts of Chemical Research 2017, 50, 517--521

  2. [2]

    Catalysis: Concepts and Green Applications; Wiley-VCH: Weinheim, Germany, 2008

    Rothenberg, G. Catalysis: Concepts and Green Applications; Wiley-VCH: Weinheim, Germany, 2008

  3. [3]

    Nanoscale engineering of catalytic materials for sustainable technologies

    Mitchell, S.; Qin, R.; Zheng, N.; Pérez-Ramírez, J. Nanoscale engineering of catalytic materials for sustainable technologies. Nature Nanotechnology 2021, 16, 129--139

  4. [4]

    S.; Kühl, S.; Strasser, P.; Cuenya, B

    Mistry, H.; Varela, A. S.; Kühl, S.; Strasser, P.; Cuenya, B. R. Nanostructured electrocatalysts with tunable activity and selectivity. Nature Reviews Materials 2016, 1

  5. [5]

    A.; Goldsmith, B

    Esterhuizen, J. A.; Goldsmith, B. R.; Linic, S. Interpretable machine learning for knowledge generation in heterogeneous catalysis. Nature Catalysis 2022, 5, 175--184

  6. [6]

    M.; Girgsdies, F.; Hashagen, M.; Kube, P.; Hävecker, M.; Carey, S

    Foppa, L.; Ghiringhelli, L. M.; Girgsdies, F.; Hashagen, M.; Kube, P.; Hävecker, M.; Carey, S. J.; Tarasov, A.; Kraus, P.; Rosowski, F.; Schlögl, R.; Trunschke, A.; Scheffler, M. Materials genes of heterogeneous catalysis from clean experiments and artificial intelligence. MRS Bulletin 2021, 46, 1016--1026, Open access, Impact Article, Published: 01 October 2021

  7. [7]

    A.; Jorner, K.; Žarko Ivković; López, N.; Aspuru-Guzik, A

    Pablo-García, S.; Morandi, S.; Vargas-Hernández, R. A.; Jorner, K.; Žarko Ivković; López, N.; Aspuru-Guzik, A. Fast evaluation of the adsorption energy of organic molecules on metals via graph neural networks. Nature Computational Science 2023, 3, 433--442

  8. [8]

    A multi-modal transformer for predicting global minimum adsorption energy

    Chen, J.; Huang, X.; Hua, C.; He, Y.; Schwaller, P. A multi-modal transformer for predicting global minimum adsorption energy. Nature Communications 2025, 16, 3232

Show all 27 references
  1. [9]

    M.; Wander, B.; Das, A.; Uyttendaele, M.; Zitnick, C

    Lan, J.; Palizhati, A.; Shuaibi, M.; Wood, B. M.; Wander, B.; Das, A.; Uyttendaele, M.; Zitnick, C. L.; Ulissi, Z. W. AdsorbML: a leap in efficiency for adsorption energy calculations using generalizable machine learning potentials. npj Computational Materials 2023, 9, 172

  2. [10]

    M.; Kley, K

    Mauß, J. M.; Kley, K. S.; Khobragade, R.; Tran, N.-K.; Bellis, J. D.; Schüth, F.; Scheffler, M.; Foppa, L. Modelling the Time-Dependent Reactivity of Catalysts by Experiments and Artificial Intelligence. ChemRxiv 2025, Preprint, not peer-reviewed

  3. [11]

    Exploring the Structure–Activity Relationship on Platinum Nanoparticles

    Braunwarth, L.; Jung, C.; Jacob, T. Exploring the Structure–Activity Relationship on Platinum Nanoparticles. Topics in Catalysis 2020, 63, 1647--1657

  4. [12]

    G.; Baletto, F

    Rossi, K.; Asara, G. G.; Baletto, F. Structural Screening and Design of Platinum Nanosamples for Oxygen Reduction. ACS Catalysis 2020, 10, 3520--3533

  5. [13]

    Calle-Vallejo, F.; Loffreda, D.; Koper, M. T. M.; Sautet, P. Introducing structural sensitivity into adsorption–energy scaling relations by means of coordination numbers. Nature Chemistry 2015, 7, 403--410

  6. [14]

    H.; Pohl, M

    Calle-Vallejo, F.; Tymoczko, J.; Colic, V.; Vu, Q. H.; Pohl, M. D.; Morgenstern, K.; Loffreda, D.; Sautet, P.; Schuhmann, W.; Bandarenka, A. S. Finding optimal surface sites on heterogeneous catalysts by counting nearest neighbors. Science 2015, 350, 185--189

  7. [15]

    E.; Williams, C

    Rasmussen, C. E.; Williams, C. K. Gaussian Processes for Machine Learning; MIT Press: Cambridge, MA, 2006

  8. [16]

    Frazier, P. I. A tutorial on Bayesian optimization. arXiv preprint arXiv:1807.02811 2018,

  9. [17]

    D.; Reinisch, D.; Loffreda, D.; Sautet, P.; Bandarenka, A

    Calle‐Vallejo, F.; Pohl, M. D.; Reinisch, D.; Loffreda, D.; Sautet, P.; Bandarenka, A. S. Why conclusions from platinum model surfaces do not necessarily lead to enhanced nanoparticle catalysts for the oxygen reduction reaction. Chemical Science 2017, 8, 2283--2289

  10. [18]

    W.; Billinge, S

    Terban, M. W.; Billinge, S. J. L. Structural Analysis of Molecular Materials Using the Pair Distribution Function. Chemical Reviews 2022, 122, 1208--1272

  11. [19]

    A universal signature in the melting of metallic nanoparticles

    Delgado-Callico, L.; Rossi, K.; Pinto-Miles, R.; Salzbrenner, P.; Baletto, F. A universal signature in the melting of metallic nanoparticles. Nanoscale 2021, 13, 1172--1180

  12. [20]

    Charting Nanocluster Structures via Convolutional Neural Networks

    Telari, E.; Tinti, A.; Settem, M.; Maragliano, L.; Ferrando, R.; Giacomello, A. Charting Nanocluster Structures via Convolutional Neural Networks. ACS Nano 2023, 17, 21287--21296

  13. [21]

    Beyond Strain and Ligand Effects: Microstrain-Induced Enhancement of the Oxygen Reduction Reaction Kinetics on Various PtNi/C Nanostructures

    Chattot, R.; Asset, T.; Bordet, P.; Drnec, J.; Dubau, L.; Maillard, F. Beyond Strain and Ligand Effects: Microstrain-Induced Enhancement of the Oxygen Reduction Reaction Kinetics on Various PtNi/C Nanostructures. ACS Catalysis 2017, 7, 398--408

  14. [22]

    Building Practical Descriptors for Defect Engineering of Electrocatalytic Materials

    Chattot, R.; Bordet, P.; Martens, I.; Drnec, J.; Dubau, L.; Maillard, F. Building Practical Descriptors for Defect Engineering of Electrocatalytic Materials. ACS Catalysis 2020, 10, 9046--9056

  15. [23]

    Timoshenko, J.; Lu, D.; Lin, Y.; Frenkel, A. I. Supervised Machine-Learning-Based Determination of Three-Dimensional Structure of Metallic Nanoparticles. The Journal of Physical Chemistry Letters 2017, 8, 5091--5098

  16. [24]

    M.; Seeger, M

    Srinivas, N.; Krause, A.; Kakade, S. M.; Seeger, M. W. Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting. IEEE Transactions on Information Theory 2012, 58, 3250--3265

  17. [25]

    Quantitatively Determining Surface–Adsorbate Properties from Vibrational Spectroscopy with Interpretable Machine Learning

    Wang, X.; Jiang, S.; Hu, W.; Ye, S.; Wang, T.; Wu, F.; Yang, L.; Li, X.; Zhang, G.; Chen, X.; Jiang, J.; Luo, Y. Quantitatively Determining Surface–Adsorbate Properties from Vibrational Spectroscopy with Interpretable Machine Learning. Journal of the American Chemical Society ...

  18. [26]

    L.; Vlachos, D

    Lansford, J. L.; Vlachos, D. G. Infrared spectroscopy data- and physics-driven machine learning for characterizing surface microstructure of complex materials. Nature Communications 2020, 11, 1513

  19. [27]

    -ݢ"! B AQD@ i| fa 7vuwO^;wf @ @ @ @ @ @ @ @ ^ [jל\* #^H jל .A x . ) X_s2,+8 / 5+X H P_| bK6

    Anker, A. S.; Jensen, J. H.; Gonzalez-Duque, M.; Moreno, R.; Smolska, A.; Juelsholt, M.; Hardion, V.; Jorgensen, M. R. V.; Faina, A.; Quinson, J.; Stoy, K.; Vegge, T. Autonomous nanoparticle synthesis by design. arXiv preprint arXiv:2505.13571 2025, mcitethebibliography acs-ma...

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.