Pith. sign in

REVIEW 4 major objections 3 minor 57 references

Neural Operators for Forward and Inverse Potential-Density Mappings in Classical Density Functional Theory

T0 review · 4 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A Fourier neural operator learns the density-to-correlation map of a hard-rod fluid more accurately than DeepONet or dense networks, and generalizes better to unseen potentials.

desk verdict A credible FNO benchmark for cDFT, but the headline extrapolation test doesn't actually test extrapolation. read the letter →

arxiv 2506.06623 v2 pith:B2P6GVRJ submitted 2025-06-07 physics.chem-ph math.STphysics.comp-phstat.TH

Runtong Pan , Xinyi Fang , Kamyar Azizzadenesheli , Miguel Liu-Schiaffini , Mengyang Gu , Jianzhong Wu This is my paper
classification physics.chem-phmath.STphysics.comp-phstat.TH
keywords neuraloperatorsclassicaldensityfunctionaltheoryFourieroperatorDeepONetdirectcorrelationfunctionhard-rodfluidsextrapolationscalinglaw
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

This paper asks whether neural operator architectures, which learn maps between whole functions rather than between fixed-length vectors, can supply the functional relationships that classical density functional theory (cDFT) needs, specifically the map from a fluid's density profile ρ(x) to its one-body direct correlation function c₁(x). Using exact analytical solutions for one-dimensional hard-rod fluids as training data, the authors compare eight machine-learning models—several DeepONet variants, a Fourier neural operator (FNO), and dense-network baselines—and report that FNO is the most accurate on every metric, including extrapolation to a held-out potential family. They further establish that a learned ρ ↦ c₁ operator, when inserted into the Euler–Lagrange equation, reproduces density profiles, whereas a directly learned potential-to-density mapping fails out of distribution. A scaling-law analysis is used to argue that operator learning has a lower irreducible error than discrete functional learning and therefore gains more from added data and compute. If right, this makes FNO a practical route to data-driven free-energy functionals for systems where no accurate analytic functional exists.

What carries the argument

The load-bearing object is the operator $G_c : \rho(x) \mapsto c_1(x)$, an infinite-dimensional map approximated by a Fourier neural operator. FNO works by lifting the discretized input into multiple channels, passing them through Fourier layers in which a truncated fast Fourier transform multiplies the spectral coefficients by a trainable weight matrix before an inverse transform and nonlinear activation, then projecting back to the output function; the spectral multiplication acts as a global convolution, which is how the network captures the nonlocal functional dependence of $c_1$ on $\rho$. The squared-ReLU activation $\sigma(x) = x(2-x)$ for $0 < x < 2$, compactly supported and with non-negative derivatives, is identified as the best nonlinearity for extrapolation. Excess free energy is recovered by thermodynamic integration of $c_1$ along a density-scaling path, and density profiles are obtained by solving the Euler–Lagrange equation with the learned $c_1$, a pipeline the paper calls operator-DFT.

What would settle it

Withhold an entire potential family during training—for example, exclude the constant-potential Group I or the power-law Group IV and test on it—or generate a truly novel family such as sinusoidal wells, then compare FNO against GK-RMSCNN-DeepONet and the quasi-local DNN on that set. If FNO's error advantage narrows to within one order of magnitude, the claim that operator learning generalizes far better than discrete functional learning loses its support.

Watch

Extended reading notes

Core claim

The central claim is that a Fourier neural operator learns the density-to-correlation functional $G_c : \rho(x) \mapsto c_1(x)$ for a one-dimensional hard-rod fluid accurately enough to serve as the free-energy engine of a density functional calculation, and that it does so clearly better than DeepONet variants and dense neural networks. On the analytic reference data, FNO reaches a mean-squared error of about $5 \times 10^{-7}$ on in-group tests and $7.8 \times 10^{-6}$ on the held-out 'new data' set, roughly three orders of magnitude below the best DeepONet variant on extrapolation. When the learned map is inserted into the Euler–Lagrange equation, the predicted density profiles track the analytical solution, while a neural operator trained directly on the map $V_{\rm ext} \mapsto \rho$ shows much larger extrapolation error, so the paper concludes that the $\rho \mapsto c_1$ route is the sound way to build operator-DFT. Free energies computed by thermodynamic integration of the learned $c_1$ reach $R^2 \approx 0.999$ on all test sets, and the Chinchilla scaling law fits every method, with FNO showing the lowest irreducible loss among them.

Load-bearing premise

The extrapolation verdict rests on a single held-out potential family, Group V, which is a linear combination of the Group II and Group III potentials and therefore lies partly inside the range the models have already seen; if a genuinely different potential class is encountered, FNO's measured extrapolation advantage could shrink.

Editorial extensions

If this is right

  • FNO is the best of the eight architectures for the $\rho \mapsto c_1$ map on both interpolation and extrapolation metrics, with held-out MSE near $8 \times 10^{-6}$ compared with $10^{-3}$–$10^{-2}$ for the dense-network baselines.
  • Squared ReLU consistently outperforms ReLU, SeLU, and logistic activations for out-of-distribution prediction in nearly every model, making it the default activation choice for operator-DFT.
  • The $\rho \mapsto c_1$ operator, combined with solving the Euler–Lagrange equation, yields density profiles competitive with the numerical solver, and it serves as a fast sample generator for active-learning Gaussian-process surrogates.
  • Chinchilla scaling-law fits predict an irreducible new-data MSE of about $2.5 \times 10^{-6}$ for FNO at infinite data and model size, more than two orders of magnitude below the fitted limits of the full-scale DNN and DeepONet baselines.

Reading between the lines

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

  • The full-range versus quasi-local distinction suggests a natural stress test: near a critical point, where the correlation length diverges, quasi-local functionals break down; whether FNO's global convolutions keep their accuracy there is an open question this hard-rod benchmark cannot answer.
  • Because FNO's trainable parameters live in channels and Fourier modes rather than on the spatial grid, its near-flat error across box sizes $L = 4$ to $12$ hints that one trained operator could be reused at different system sizes with minimal retraining.
  • Training the operator directly on the integrated free energy, rather than on pointwise $c_1$ values, might sharpen free-energy predictions further; the paper's own results show $F_{\rm ex}$ is predicted more accurately than the underlying $c_1$ field, indicating that integration averages out errors.
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 / 3 minor

Summary. This paper evaluates several neural operator architectures (DeepONet variants and FNO) for learning the mapping from one-dimensional density profiles to the one-body direct correlation function in classical density functional theory, using analytical hard-rod data as reference. It also examines the direct Vext-to-rho operator, operator-DFT solvers built from the learned rho-to-c1 map, and ALEC-GPR active learning. The models are compared through MSE losses on an in-group test set and a 'new data' test set, excess free-energy predictions, a Chinchilla scaling-law analysis, and transfer experiments across box sizes and to a Lennard-Jones fluid. The main claims are that FNO achieves the best overall accuracy, that squared ReLU is the best activation function, and that operator learning generalizes better than discrete ML approaches in extrapolation.

Significance. If the central claims were fully established, the paper would provide a useful benchmark for neural operators in cDFT: the rho-to-c1 task is validated against the exact Percus/Vanderlick analytical solution, excess free energies are compared with exact hard-rod results, and FNO is tested on multiple system sizes and on an LJ fluid including a phase-separation scenario. The negative result for the direct Vext-to-rho operator is honest and instructive, and the data/code availability statement is a strength. However, the extrapolation claim that distinguishes operator learning from discrete ML is not yet supported by the test design, and the single-run results provide no uncertainty quantification. These issues are fixable but require additional experiments and careful re-analysis.

major comments (4)
  1. [II D, Appendix B (Eq. B.5)] The 'new data' extrapolation set is Group V, defined as beta Vext(x) = mg x - epsilon[(dhr/x)^3 + (dhr/(L-x))^3] with the same parameter ranges used in Groups II and III. This is a linear combination of the training families, so it lies in the convex hull of the training inputs rather than representing a genuine distribution shift. Figure 2(a-c) indeed shows Group V t-SNE embeddings overlapping Groups II and III. Consequently, the c1-MSEND values in Table I and the fitted 'New Data' L1 values in Table II do not measure out-of-distribution generalization; they measure interpolation among known functional forms. Since the abstract and Section IV base the 'extrapolation' conclusions on this test, the claim that FNO generalizes better than discrete ML is overstrong. Please add a held-out potential family that is not expressible as a sum of the training families (e.g., an oscillatory or localized barrier) and report the new-data results for that family.
  2. [II D] The Z-score normalization computes the mean and standard deviation across all five groups, including the Group V test set. This allows test-set statistics to enter the preprocessing of the training pipeline, which can only improve the apparent new-data performance and weakens the extrapolation analysis. Normalization statistics should be computed from the training groups only (Groups I-IV) and then applied to any held-out set. This issue affects every reported 'new data' MSE and the scaling-law fits in Table II.
  3. [Tables I-V] All reported MSE values are single-run point estimates with no indication of seed variation. Because the paper's headline is a ranking of architectures (FNO best; GK-RMSCNN-DeepONet best among DeepONet variants), the results should be accompanied by means and standard deviations over several random initializations, or at least by a statement that the observed gaps are stable across seeds. Without this, Tables I and IV do not establish that the performance differences are reliable, especially for the smaller gaps among the DeepONet variants.
  4. [III C, Table II] The Chinchilla scaling-law analysis concludes that FNO has the lowest irreducible loss L1 and that operator learning generalizes better than discrete ML. This conclusion inherits the Group V test-set flaw: the 'New Data' column is fitted to the same hybrid test set, and the normalization leakage also affects these fits. The claim should be restricted to in-group generalization, or the analysis should be re-run on a properly held-out potential family before being used to support the generalization argument.
minor comments (3)
  1. [II B] The text says the list of abbreviations and inputs is 'listed in the beginning of the paper,' but the list actually appears at the end after the references; please update the reference to its actual location.
  2. [III B] The statement that there is no 'new data' test set result for the quasi-local DNN method is contradicted by Table I, which lists c1-MSEND = 7.8e-3 for that model, and by Figure 7(d), which shows New Data curves for the quasi-local DNN; please reconcile the text with the table and figure.
  3. [Various] Several typos and formatting issues remain throughout, including 'convience', 'fir', 'preditions', 'L-GBFS' (should be L-BFGS), and missing spaces in table captions such as 'Variation ofMSE'; a careful proofreading pass is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the central rho-to-c1 and Fex results are benchmarked against exact analytical solutions, and the scaling-law and activation analyses are empirical fits or selections rather than definitional reductions.

full rationale

The paper's central derivation chain is externally grounded: the operators rho->c1 and Vloc->rho are trained on data generated from the analytical 1D hard-rod solutions (Appendix A, refs 28 and 33), and all reported MSE values, Fex values (Eqs. III.4-III.5), and density profiles are compared against those exact analytical results. No fitted parameter is defined in terms of the target output, and no prediction is equivalent by construction to its training input. The Chinchilla analysis (Section III C, Table II) fits L1, A, B, alpha, and beta to the same loss curves and then uses the fitted L1 to rank asymptotic generalization; this is an in-sample extrapolation and is consistent with the observed new-data MSE in Table I, but it is a modeling inference rather than an independent prediction, so it is not a circular step in the derivation. The sRelu activation was selected after evaluating the same test sets (Sections III B and III E), which is a selection-on-test-set validity concern, not a definitional circularity. The self-citations (refs 9, 22, 24) are related-work support or are independently re-evaluated here (ALEC-GPR is benchmarked against the analytical/numerical solver in Table V), so they are not load-bearing. Two caveats weaken the 'extrapolation' label but are not circularity: Group V is explicitly a linear combination of Groups II and III (Eq. B.5) with t-SNE overlap acknowledged (Fig. 2), so leave-one-group CV is closer to interpolation between known families; and Z-score normalization uses statistics from all five groups including the test set (Section II D). These are external-validity limitations, not reductions of the results to their inputs.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the exactness of the analytical 1D hard-rod data, the cDFT framework, and the design of the extrapolation test. The fitted scaling-law constants and the hand-chosen kernel cutoff are the only tunable numbers that enter the generalization analysis; the neural network weights themselves are trained parameters and are not listed here.

free parameters (6)
  • Chinchilla scaling exponent alpha (FNO, in-group) = 0.36
    Fitted to Eq. III.2 loss curves in Section III C; used to estimate asymptotic generalization of FNO.
  • Chinchilla scaling exponent beta (FNO, in-group) = 0.26
    Fitted to Eq. III.2 loss curves in Section III C; used to estimate asymptotic generalization of FNO.
  • Chinchilla prefactor A (FNO, in-group) = 19.3
    Fitted prefactor in the scaling law Eq. III.2.
  • Chinchilla prefactor B (FNO, in-group) = 10.91
    Fitted prefactor in the scaling law Eq. III.2.
  • Irreducible loss exp(L1) (FNO, in-group) = 1.73e-7
    Asymptotic MSE of an ideal model inferred from the same loss data; used to rank methods in Table II.
  • Gaussian kernel cutoff x_conv,cut = 0.1 (10 grid points)
    Hand-chosen in Section II C to control computational cost for GK-CNN-DeepONet; affects its accuracy.
assumptions (4)
  • standard math Exactness of the analytical 1D hard-rod solutions (Percus 1988; Vanderlick et al. 1986)
    All training and reference data are generated from Eqs. A.1-A.6; if the analytical model were wrong, every benchmark target would be wrong.
  • domain assumption cDFT uniqueness: density profile determines external potential and direct correlation function (Hohenberg-Kohn-Mermin)
    Justifies learning the operators Gc: rho->c1 and G_rho: Vext->rho in Section II A.
  • domain assumption Euler-Lagrange equation (Eq. II.1) connects rho, Vloc, and c1
    The operator-DFT solvers in Section III F solve this equation; if it did not hold, the indirect approach would be invalid.
  • ad hoc to paper Group V is a meaningful out-of-distribution extrapolation test
    Group V is a linear combination of Groups II and III (Appendix B); the paper uses it as the extrapolation test set, which is a design assumption specific to this work.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural Operators for Forward and Inverse Potential-Density Mappings in Classical Density Functional Theory." pith.science (2026). https://pith.science/paper/B2P6GVRJ

@misc{pith2026250606623,
  author       = {Pith},
  title        = {Pith review of: Neural Operators for Forward and Inverse Potential-Density Mappings in Classical Density Functional Theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B2P6GVRJ}},
  note         = {Machine review of arXiv:2506.06623}
}
abstract

Neural operators are capable of capturing nonlinear mappings between infinite-dimensional functional spaces, offering a data-driven approach to modeling complex functional relationships in classical density functional theory (cDFT). In this work, we evaluate the performance of several neural operator architectures in learning the functional relationships between the one-body density profile $\rho(x)$, the one-body direct correlation function $c_1(x)$, and the external potential $V_{ext}(x)$ of inhomogeneous one-dimensional (1D) hard-rod fluids, using training data generated from analytical solutions of the underlying statistical-mechanical model. We compared their performance in terms of the Mean Squared Error (MSE) loss in establishing the functional relationships as well as in predicting the excess free energy across two test sets: (1) a group test set generated via random cross-validation (CV) to assess interpolation capability, and (2) a newly constructed dataset for leave-one-group CV to evaluate extrapolation performance. Our results show that FNO achieves the most accurate predictions of the excess free energy, with the squared ReLU activation function outperforming other activation choices. Among the DeepONet variants, the Residual Multiscale Convolutional Neural Network (RMSCNN) combined with a trainable Gaussian derivative kernel (GK-RMSCNN-DeepONet) demonstrates the best performance. Additionally, we applied the trained models to solve for the density profiles at various external potentials and compared the results with those obtained from the direct mapping $V_{ext} \mapsto \rho$ with neural operators, as well as with Gaussian Process Regression (GPR) combined with Active Learning by Error Control (ALEC), which has shown strong performance in previous studies.

Figures

Figures reproduced from arXiv: 2506.06623 by the authors.

Figure 1
Figure 1. Conceptual diagram illustrating application of neural op￾erators in cDFT [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. t-SNE visualization of (a) the reduced background po￾tential βVloc(x), (b) the one-body density profile ρ(x), and (c) the one-body direct correlation function c1(x). The t-SNE plots were generated using the Barnes-Hut algorithm with the following param￾eters: perplexity = 25, learning rate = 100, and maximum number of iterations = 1000.(d) Data split diagram illustrating the training set, in-group test set, and new … view at source ↗
Figure 3
Figure 3. Schematic of the general DeepONet architecture. The branch net consists of p unstacked sub-networks, each processing components of the discretized input function. The trunk net is a sin￾gle stacked network that produces p outputs corresponding to the sensor locations of the output functions. (RMSCNN) (multi-scale CNN with Residual neural network (ResNet)). These network designs have been widely ap￾plied to computer … view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Schematic structures of Fourier neural operators (FNO). (a) A FNO model typically consists of three main layers: a lifting layer, iterative Fourier layers (or kernel integration layers), and a projection layer. In the lifting layer, the input network P ‘lifts’ the orig…
Figure 5
Figure 5. Figure 5: shows the relation between the number of train￾ing iterations and the loss values (MSE) on the training set, the in-group test set, and the ‘new data’ test set. We have no ‘new data’ test set result for the quasi-local DNN method, as its training framework does not acc…
Figure 6
Figure 6. Figure 6: A comparison of c1(x) predicted by various ML meth￾ods, including full-ranged and quasi-local DNNs, DNN-DeepONet, GK-RMSCNN-DeepONet, FNO. The preditions of ML methods was labeled as "Pred", DeepONet labeled as "DON" (ac) Three examples from the in-group test set. (df)…
Figure 7
Figure 7. Figure 7: MSE loss versus training data size (Nsample) on the train￾ing set, in-group test set, and new data test set, for different ML methods. Training was stopped when the in-group test loss failed to decrease for 2000 steps or after reaching 100,000 steps for both DeepONet a…
Figure 8
Figure 8. Figure 8: Comparison of MSE loss versus model size Dmodel for different ML methods. Training was terminated when the in-group test loss did not decrease for 2000 steps or when 100,000 steps were reached, for both DeepONet and DNN. DNN-DeepONet is shown in place of GK-RMSCNN-Deep…
Figure 9
Figure 9. Figure 9: Comparison of the excess Helmholtz energy Fex pre￾dicted by different ML methods with the exact results. (a) FNO (b) Full-scale DNN(c) DNN-DeepONet (d) GK-RMSCNN-DeepONet (e) Quasi-local DNN. Legends are: Training Set: Blue; In-Group Test Set MSE: Orange; New data Test…
Figure 11
Figure 11. Figure 11: (a) shows that the direct correlation operator Gc achieves consistently low losses for both interpolation and ex￾trapolation across a range of system sizes, demonstrating ro￾bust predictive capability. The MSE losses are slightly higher for smaller box sizes, likely d…
Figure 12
Figure 12. Figure 12: Performance of FNO in learning the direct correlation operator Gc for one-dimensional LJ fluids confined between hard walls at system size L = 8. (a) Representative prediction of the di￾rect correlation function c1(x), demonstrating the models accuracy in capturing sp…
Figure 13
Figure 13. Figure 13: Function prediction by operator-DFT models. (a, b) Significance-relevant predictions by GK-RMSCNN-DeepONet. (c, d) Resolution migration of FNO models from high-resolution data (dz = 0.01) to low-resolution data (dz = 0.05). (a, c) Sample from data group III; (b, d) Sa…
Figure 14
Figure 14. Figure 14: Illustrative examples for the density profiles ob￾tained from operator-DFT with operator-based functions, compared against analytical results and direct operator-DFT calculations. (a) Significance-relevant prediction by GK-RMSCNN-DeepONet. (b) Resolution migration of …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

57 extracted references · 28 canonical work pages

  1. [1]

    merlin.mbs aapmrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs aapmrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translat...

    2010

  2. [2]

    merlin.mbs aipauth4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs aipauth4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translat...

    2010

  3. [3]

    merlin.mbs aipnum4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs aipnum4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translati...

    2010

  4. [4]

    merlin.mbs apsrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs apsrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translati...

    2010

  5. [5]

    merlin.mbs apsrmp4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs apsrmp4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translati...

    2010

  6. [6]

    author author R. Evans ,\ 10.1080/00018737900101365 journal journal Advances in Physics \ volume 28 ,\ pages 143 ( year 1979 ) ,\ http://arxiv.org/abs/https://doi.org/10.1080/00018737900101365 https://doi.org/10.1080/00018737900101365 NoStop

    doi:10.1080/00018737900101365 1979

  7. [7]

    Jeanmairet , author N

    author author G. Jeanmairet , author N. Levy , author M. Levesque , \ and\ author D. Borgis ,\ 10.1021/ed500049m journal journal Journal of Chemical Education \ volume 91 ,\ pages 2112 ( year 2014 ) ,\ http://arxiv.org/abs/https://doi.org/10.1021/ed500049m https://doi.org/10.1021/ed500049m NoStop

    doi:10.1021/ed500049m 2014

  8. [8]

    author author J. Wu ,\ https://link.springer.com/book/10.1007/978-981-10-2502-0 title Variational Methods in Molecular Modeling \ ( publisher Springer Berlin Heidelberg ,\ address New York, NY ,\ year 2017 )\ p.\ pages 324 NoStop

    work page doi:10.1007/978-981-10-2502-0 2017

Show all 57 references
  1. [9]

    author author J. Wu ,\ https://doi.org/10.1002/aic.10713 journal journal AIChE Journal \ volume 52 ,\ pages 1169 ( year 2006 ) ,\ http://arxiv.org/abs/https://aiche.onlinelibrary.wiley.com/doi/pdf/10.1002/aic.10713 https://aiche.onlinelibrary.wiley.com/doi/pdf/10.1002/aic.10713 NoStop

    2006 doi

  2. [10]

    Wu \ and\ author Z

    author author J. Wu \ and\ author Z. Li ,\ 10.1146/annurev.physchem.58.032806.104650 journal journal Annual Review of Physical Chemistry \ volume 58 ,\ pages 85 ( year 2007 ) ,\ http://arxiv.org/abs/https://doi.org/10.1146/annurev.physchem.58.032806.104650 https://doi.org/10.1...

    2007

  3. [11]

    de Morais Sermoud , author A

    author author V. de Morais Sermoud , author A. de Freitas Gonçalves , author A. G. \ Barreto Jr. , author L. F. M. \ Franco , author F. W. \ Tavares , \ and\ author M. Castier ,\ https://doi.org/10.1016/j.fluid.2024.114177 journal journal Fluid Phase Equilibria \ volume 586 ,\...

    2024

  4. [12]

    Sammüller , author S

    author author F. Sammüller , author S. Hermann , \ and\ author M. Schmidt ,\ 10.1088/1361-648X/ad326f journal journal Journal of Physics: Condensed Matter \ volume 36 ,\ pages 243002 ( year 2024 ) NoStop

    2024 doi

  5. [13]

    Simon \ and\ author M

    author author A. Simon \ and\ author M. Oettel ,\ https://arxiv.org/abs/2406.07345 title Machine learning approaches to classical density functional theory , \ ( year 2024 ),\ http://arxiv.org/abs/2406.07345 arXiv:2406.07345 [cond-mat.stat-mech] NoStop

    2024 arXiv

  6. [14]

    Wu \ and\ author M

    author author J. Wu \ and\ author M. Gu ,\ 10.1021/acs.jpclett.3c02804 journal journal The Journal of Physical Chemistry Letters \ volume 14 ,\ pages 10545 ( year 2023 ) ,\ http://arxiv.org/abs/https://doi.org/10.1021/acs.jpclett.3c02804 https://doi.org/10.1021/acs.jpclett.3c0...

    2023 doi

  7. [15]

    \ Lin \ and\ author M

    author author S.-C. \ Lin \ and\ author M. Oettel ,\ 10.21468/SciPostPhys.6.2.025 journal journal SciPost Phys. \ volume 6 ,\ pages 025 ( year 2019 ) NoStop

    2019 doi

  8. [16]

    \ Lin , author G

    author author S.-C. \ Lin , author G. Martius , \ and\ author M. Oettel ,\ 10.1063/1.5135919 journal journal The Journal of Chemical Physics \ volume 152 ,\ pages 021102 ( year 2020 ) ,\ http://arxiv.org/abs/https://pubs.aip.org/aip/jcp/article-pdf/doi/10.1063/1.5135919/155686...

    2020 doi

  9. [17]

    author author J. V. \ Dillon , author I. Langmore , author D. Tran , author E. Brevdo , author S. Vasudevan , author D. Moore , author B. Patton , author A. Alemi , author M. Hoffman , \ and\ author R. A. \ Saurous ,\ 10.48550/arXiv.1711.10604 journal journal arXiv e-prints \ ...

    doi arXiv

  10. [18]

    author author M. M. \ Kelley , author J. Quinton , author K. Fazel , author N. Karimitari , author C. Sutton , \ and\ author R. Sundararaman ,\ 10.1063/5.0223792 journal journal The Journal of Chemical Physics \ volume 161 ,\ pages 144101 ( year 2024 ) ,\ http://arxiv.org/abs/...

    doi

  11. [19]

    Malpica-Morales , author P

    author author A. Malpica-Morales , author P. Yatsyshin , author M. A. \ Durán-Olivencia , \ and\ author S. Kalliadasis ,\ 10.1063/5.0146920 journal journal The Journal of Chemical Physics \ volume 159 ,\ pages 104109 ( year 2023 ) ,\ http://arxiv.org/abs/https://pubs.aip.org/a...

    doi

  12. [20]

    Yatsyshin , author S

    author author P. Yatsyshin , author S. Kalliadasis , \ and\ author A. B. \ Duncan ,\ 10.1063/5.0071629 journal journal The Journal of Chemical Physics \ volume 156 ,\ pages 074105 ( year 2022 ) ,\ http://arxiv.org/abs/https://pubs.aip.org/aip/jcp/article-pdf/doi/10.1063/5.0071...

    2022 doi

  13. [21]

    Cats , author S

    author author P. Cats , author S. Kuipers , author S. de Wind , author R. van Damme , author G. M. \ Coli , author M. Dijkstra , \ and\ author R. van Roij ,\ 10.1063/5.0042558 journal journal APL Materials \ volume 9 ,\ pages 031109 ( year 2021 ) ,\ http://arxiv.org/abs/https:...

    doi

  14. [22]

    Dijkman , author M

    author author J. Dijkman , author M. Dijkstra , author R. van Roij , author M. Welling , author J.-W. \ van de Meent , \ and\ author B. Ensing ,\ 10.1103/PhysRevLett.134.056103 journal journal Phys. Rev. Lett. \ volume 134 ,\ pages 056103 ( year 2025 ) NoStop

    doi

  15. [24]

    Sammüller , author S

    author author F. Sammüller , author S. Hermann , author D. de Las Heras , \ and\ author M. Schmidt ,\ 10.1073/pnas.2312484120 journal journal Proc Natl Acad Sci U S A \ volume 120 ,\ pages e2312484120 ( year 2023 ) NoStop

    2023 doi

  16. [25]

    Samm\"uller , author M

    author author F. Samm\"uller , author M. Schmidt , \ and\ author R. Evans ,\ 10.1103/PhysRevX.15.011013 journal journal Phys. Rev. X \ volume 15 ,\ pages 011013 ( year 2025 ) NoStop

    2025 doi

  17. [26]

    author author A. T. \ Bui \ and\ author S. J. \ Cox ,\ 10.1103/PhysRevLett.134.148001 journal journal Phys. Rev. Lett. \ volume 134 ,\ pages 148001 ( year 2025 ) NoStop

    2025 doi

  18. [27]

    Yang , author R

    author author J. Yang , author R. Pan , author J. Sun , \ and\ author J. Wu ,\ 10.1021/acs.jctc.5c00484 journal journal Journal of Chemical Theory and Computation \ volume 0 ,\ pages null ( year 0 ) ,\ http://arxiv.org/abs/https://doi.org/10.1021/acs.jctc.5c00484 https://doi.o...

    doi

  19. [28]

    Santos-Silva , author P

    author author T. Santos-Silva , author P. I. C. \ Teixeira , author C. Anquetil-Deck , \ and\ author D. J. \ Cleaver ,\ 10.1103/PhysRevE.89.053316 journal journal Phys. Rev. E \ volume 89 ,\ pages 053316 ( year 2014 ) NoStop

    2014 doi

  20. [29]

    Fang , author M

    author author X. Fang , author M. Gu , \ and\ author J. Wu ,\ 10.1063/5.0121805 journal journal J Chem Phys \ volume 157 ,\ pages 214109 ( year 2022 ) NoStop

    2022 doi

  21. [31]

    Wang , author T

    author author H. Wang , author T. Fu , author Y. Du , author W. Gao , author K. Huang , author Z. Liu , author P. Chandak , author S. Liu , author P. Van Katwyk , author A. Deac , et al. ,\ @noop journal journal Nature \ volume 620 ,\ pages 47 ( year 2023 ) NoStop

    2023

  22. [32]

    Kovachki , author Z

    author author N. Kovachki , author Z. Li , author B. Liu , author K. Azizzadenesheli , author K. Bhattacharya , author A. Stuart , \ and\ author A. Anandkumar ,\ http://jmlr.org/papers/v24/21-1524.html journal journal Journal of Machine Learning Research \ volume 24 ,\ pages 1...

    2023

  23. [33]

    author author T. K. \ Vanderlick , author L. E. \ Scriven , \ and\ author H. T. \ Davis ,\ 10.1103/physreva.34.5130 journal journal Phys Rev A Gen Phys \ volume 34 ,\ pages 5130 ( year 1986 ) NoStop

    1986 doi

  24. [34]

    author author E. S. \ Muckley , author J. E. \ Saal , author B. Meredig , author C. S. \ Roper , \ and\ author J. H. \ Martin ,\ 10.1039/d3dd00082f journal journal Digital Discovery \ volume 2 ,\ pages 1425 ( year 2023 ) NoStop

    2023 doi

  25. [35]

    Kaplan , author S

    author author J. Kaplan , author S. McCandlish , author T. Henighan , author T. B. \ Brown , author B. Chess , author R. Child , author S. Gray , author A. Radford , author J. Wu , \ and\ author D. Amodei ,\ https://arxiv.org/abs/2001.08361 title Scaling laws for neural langua...

    2001 arXiv

  26. [36]

    Hohenberg \ and\ author W

    author author P. Hohenberg \ and\ author W. Kohn ,\ @noop journal journal Phys. Rev. B \ volume 136 ,\ pages B864 ( year 1964 ) NoStop

    1964

  27. [37]

    author author N. D. \ Mermin ,\ @noop journal journal Phys. Rev. \ volume 137 ,\ pages A1441 ( year 1965 ) NoStop

    1965

  28. [38]

    author author J. K. \ Percus ,\ 10.1007/bf01011639 journal journal Journal of Statistical Physics \ volume 52 ,\ pages 1157 ( year 1988 ) NoStop

    1988 doi

  29. [40]

    Li , author N

    author author Z. Li , author N. Kovachki , author K. Azizzadenesheli , author B. Liu , author K. Bhattacharya , author A. Stuart , \ and\ author A. Anandkumar ,\ https://arxiv.org/abs/2010.08895 title Fourier neural operator for parametric partial differential equations , \ ( ...

    2010 arXiv

  30. [41]

    Diab \ and\ author M

    author author W. Diab \ and\ author M. Al Kobaisi ,\ 10.1038/s41598-024-72393-0 journal journal Sci Rep \ volume 14 ,\ pages 21298 ( year 2024 ) NoStop

    2024 doi

  31. [42]

    Kontolati , author S

    author author K. Kontolati , author S. Goswami , author G. E. \ Karniadakis , \ and\ author M. D. \ Shields ,\ 10.48550/arXiv.2304.07599 title Learning in latent spaces improves the predictive accuracy of deep neural operators , \ ( year 2023 ) NoStop

    doi arXiv

  32. [43]

    Lin , author Z

    author author C. Lin , author Z. Li , author L. Lu , author S. Cai , author M. Maxey , \ and\ author G. E. \ Karniadakis ,\ 10.1063/5.0041203 journal journal J Chem Phys \ volume 154 ,\ pages 104118 ( year 2021 ) NoStop

    2021 doi

  33. [44]

    Lu , author X

    author author L. Lu , author X. Meng , author S. Cai , author Z. Mao , author S. Goswami , author Z. Zhang , \ and\ author G. E. \ Karniadakis ,\ 10.1016/j.cma.2022.114778 journal journal Computer Methods in Applied Mechanics and Engineering \ volume 393 ( year 2022 b ),\ 10.1...

    2022

  34. [45]

    author author F. I. \ Diakogiannis , author F. Waldner , author P. Caccetta , \ and\ author C. Wu ,\ 10.1016/j.isprsjprs.2020.01.013 journal journal ISPRS Journal of Photogrammetry and Remote Sensing \ volume 162 ,\ pages 94 ( year 2020 ) NoStop

    2020 doi

  35. [46]

    \ Gao , author M.-M

    author author S.-H. \ Gao , author M.-M. \ Cheng , author K. Zhao , author X.-Y. \ Zhang , author M.-H. \ Yang , \ and\ author P. Torr ,\ 10.1109/tpami.2019.2938758 journal journal IEEE Transactions on Pattern Analysis and Machine Intelligence \ volume 43 ,\ pages 652–662 ( ye...

    2019

  36. [47]

    Sosnovik , author A

    author author I. Sosnovik , author A. Moskalev , \ and\ author A. Smeulders ,\ @noop journal journal arXiv preprint arXiv:2106.02733 \ ( year 2021 ) NoStop

    2021 arXiv

  37. [48]

    Bahri , author E

    author author Y. Bahri , author E. Dyer , author J. Kaplan , author J. Lee , \ and\ author U. Sharma ,\ 10.1073/pnas.2311878121 journal journal Proceedings of the National Academy of Sciences \ volume 121 ( year 2024 ),\ 10.1073/pnas.2311878121 NoStop

    2024 doi

  38. [49]

    author author J. D. \ Wang , author J. C. \ Xu , \ and\ author J. Q. \ Zhu ,\ 10.1007/978-3-031-08754-7_40 journal journal Computational Science, Iccs 2022, Pt Ii \ ,\ pages 319 ( year 2022 ) NoStop

    2022 doi

  39. [50]

    Lu , author X

    author author L. Lu , author X. Meng , author Z. Mao , \ and\ author G. E. \ Karniadakis ,\ 10.1137/19M1274067 journal journal SIAM Review \ volume 63 ,\ pages 208 ( year 2021 ) NoStop

    2021 doi

  40. [51]

    Penaud-Polge , author S

    author author V. Penaud-Polge , author S. Velasco-Forero , \ and\ author J. Angulo ,\ 10.1109/Icip46576.2022.9897734 journal journal 2022 Ieee International Conference on Image Processing, Icip \ ,\ pages 2421 ( year 2022 ) NoStop

    2022

  41. [52]

    Li , author N

    author author Z. Li , author N. Kovachki , author K. Azizzadenesheli , author B. Liu , author K. Bhattacharya , author A. Stuart , \ and\ author A. Anandkumar ,\ 10.48550/arXiv.2010.08895 title Fourier neural operator for parametric partial differential equations , \ ( year 20...

    2010 doi arXiv

  42. [53]

    author author N. B. \ Kovachki , author Z. Li , author B. Liu , author K. Azizzadenesheli , author K. Bhattacharya , author A. M. \ Stuart , \ and\ author A. Anandkumar ,\ @noop journal journal CoRR \ volume abs/2108.08481 ( year 2021 ) NoStop

    2021 arXiv

  43. [54]

    Paszke , author S

    author author A. Paszke , author S. Gross , author F. Massa , author A. Lerer , author J. Bradbury , author G. Chanan , author T. Killeen , author Z. Lin , author N. Gimelshein , author L. Antiga , author A. Desmaison , author A. K \"o pf , author E. Yang , author Z. DeVito , ...

    doi arXiv

  44. [55]

    Bonev , author T

    author author B. Bonev , author T. Kurth , author C. Hundt , author J. Pathak , author M. Baust , author K. Kashinath , \ and\ author A. Anandkumar ,\ in\ https://proceedings.mlr.press/v202/bonev23a.html booktitle Proceedings of the 40th International Conference on Machine Lea...

    2023

  45. [56]

    Tan \ and\ author L

    author author L. Tan \ and\ author L. Chen ,\ 10.48550/arXiv.2202.08942 title Enhanced deeponet for modeling partial differential operators considering multiple input functions , \ ( year 2022 ) NoStop

    2022 doi arXiv

  46. [57]

    Jiang , author M

    author author Z. Jiang , author M. Zhu , \ and\ author L. Lu ,\ https://doi.org/10.1016/j.ress.2024.110392 journal journal Reliability Engineering & System Safety \ volume 251 ,\ pages 110392 ( year 2024 ) NoStop

    2024

  47. [58]

    Sammüller \ and\ author M

    author author F. Sammüller \ and\ author M. Schmidt ,\ ARTN L032601 10.1103/PhysRevE.110.L032601 journal journal Physical Review E \ volume 110 ( year 2024 ),\ ARTN L032601 10.1103/PhysRevE.110.L032601 NoStop

    2024 doi

  48. [59]

    Li , author H

    author author Z. Li , author H. Zheng , author N. Kovachki , author D. Jin , author H. Chen , author B. Liu , author K. Azizzadenesheli , \ and\ author A. Anandkumar ,\ 10.48550/arXiv.2111.03794 title Physics-informed neural operator for learning partial differential equations...

    doi arXiv

  49. [60]

    Goswami , author A

    author author S. Goswami , author A. Bora , author Y. Yu , \ and\ author G. E. \ Karniadakis ,\ title Physics-informed deep neural operator networks , \ in\ 10.1007/978-3-031-36644-4_6 booktitle Machine Learning in Modeling and Simulation: Methods and Applications ,\ editor ed...

    doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.