Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

The Bigger the Better? Accurate Molecular Potential Energy Surfaces from Minimalist Neural Networks

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Tiny neural networks match million-parameter molecular energy models

desk verdict A small hybrid kernel/NN potential (KerNN, ~10^3 params) matches PhysNet on accuracy for the tested systems and is much faster; the extrapolation claim is real but should be scoped more carefully than the abstract does. read the letter →

arxiv 2411.18121 v1 pith:LCSLNFGQ submitted 2024-11-27 physics.chem-ph cs.LG

Silvan Käser , Debasish Koner , Markus Meuwly This is my paper · ORCID
classification physics.chem-phcs.LG
keywords potentialenergysurfaceneuralnetworkreproducingkernelextrapolationparsimonymoleculardynamicsinfraredspectroscopyreactivescattering
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 claims that accurate molecular potential energy surfaces do not require the huge neural networks currently in use. It introduces KerNN, which feeds one-dimensional reproducing kernel functions of interatomic distances into a small feed-forward network with about a thousand parameters. On formaldehyde, HeH2+, and hydrogen oxalate, this tiny network matches or beats PhysNet—a network with roughly a million parameters—on energies, forces, dipole moments, and computed spectra, and it extrapolates more reliably beyond the training data. If correct, this would make machine-learned molecular dynamics far cheaper and more robust, especially in dissociation and bond-breaking regions. The paper's central insight is that simpler, smaller models can match or exceed larger ones when the descriptors encode the right physics.

What carries the argument

The carrying object is the one-dimensional reciprocal-power reproducing kernel k[3,3](r, r'), a smooth similarity function of two interatomic distances that decays as $r^{{-4}}$ for large arguments. It converts raw distances into features that enter a two-hidden-layer feed-forward network with softplus activations; the kernel's monotonic decay toward zero is what gives the network a sensible long-range asymptote. A symmetrized variant built from fundamental invariants imposes permutational invariance for like atoms.

What would settle it

Train KerNN on a molecule that dissociates into ions, e.g., NaCl, where the energy approaches the Coulomb law -1/R rather than a constant; if the predicted potential levels off instead of following the 1/R curve for R beyond the training data, the claimed extrapolation advantage fails.

Watch

Extended reading notes

Core claim

The central discovery is that a neural-network potential energy surface can be built from about $10^{3}$ parameters without losing accuracy. KerNN uses one-dimensional reciprocal-power reproducing kernels k[3,3](r, r') as input features, each measuring the similarity between an interatomic distance and a reference distance, and a two-hidden-layer feed-forward network with softplus activations maps those features to the total energy. On the same ab initio reference data, this small network reaches test-set errors comparable to PhysNet's $10^{6}$-parameter model for energy, forces, and dipole moments across three very different molecules. The kernel features also improve extrapolation: because k[3,3] decays monotonically toward zero for large distances, the network is smoothly anchored as bonds are stretched far beyond the training range. This is demonstrated on 5000 K-sampled geometries, one-dimensional dissociation cuts, and a transfer-learned correction to the experimental dissociation energy of formaldehyde, and on spectroscopy and reactivity for HeH2+ and hydrogen oxalate.

Load-bearing premise

The central premise is that a feed-forward network whose kernel features have saturated at zero will still produce physical energies outside the training range; this is an empirical property of the trained models, not a mathematical guarantee, so the extrapolation claim could fail for a different molecule or training set.

Editorial extensions

If this is right

  • Training and evaluation costs drop by roughly two orders of magnitude for the systems tested, making long-time molecular dynamics simulations with machine-learned potentials practical.
  • Because the kernel features decay monotonically outside the training range, the potential remains physical in dissociation and bond-breaking regions, where typical neural-network potentials drift.
  • The same tiny network can be adapted to predict dipole moments and infrared spectra, so accelerated simulations feed directly into spectroscopic observables.
  • Unlike kernel ridge regression, KerNN's evaluation cost does not grow with the number of training points, so it can scale to larger reference data sets.

Reading between the lines

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

  • A stronger version of the paper's claim would be that descriptor physics, not network capacity, sets the accuracy ceiling; this could be tested by ablating the kernel feature set against raw distances or exponentials while holding the architecture fixed.
  • The reliable long-range behavior holds for dissociations whose true asymptote is flat; for ionic or other slowly decaying long-range tails, the current k[3,3] features would saturate and the potential would plateau, so the extrapolation guarantee is specific to the tested systems.
  • The hydrogen oxalate breakdown after hydrogen-bond breaking suggests that for floppy, large-amplitude systems, unsymmetrized descriptors need data augmentation or approximate permutational invariance before the approach can be used without restrictions.
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

3 major / 5 minor

Summary. The paper introduces KerNN, a neural-network potential energy surface (PES) model that uses one-dimensional reciprocal power reproducing kernels k[3,3] as fixed input features to a small fully connected network (two hidden layers, about 10^3 parameters). KerNN is applied to formaldehyde (H2CO), HeH2+, and hydrogen oxalate. The authors report that KerNN matches or exceeds the accuracy of PhysNet (which has about 10^6 parameters) and other baselines on held-out energies and forces, harmonic frequencies, DVR3D bound states for HeH2+, IR spectra from molecular dynamics, and hydrogen-transfer barriers and rates. They also claim that the kernel features improve extrapolation far beyond the training data, citing a 5000 K extrapolation set for H2CO, one-dimensional C-H dissociation scans with transfer learning to the experimental dissociation energy, and long-range HeH2+ scans. The paper concludes that considerably smaller and simpler ML models can be competitive, and that kernel features solve a general extrapolation problem of NN-based PESs.

Significance. If the central claim holds, KerNN is a significant contribution: it demonstrates that extremely compact NN architectures with tailored fixed features can provide state-of-the-art accuracy and orders-of-magnitude speedups, and it offers a possible route to controlling asymptotic behavior via kernel descriptors. The paper benefits from multiple independent validations: out-of-sample error statistics, harmonic frequencies, quantum bound-state calculations (DVR3D), classical IR spectra, and experimental comparison for HeH2+. The transfer-learning experiment using the experimental C-H dissociation energy is a useful demonstration. However, the extrapolation claim is more general than the evidence supports, and reproducibility is limited by the absence of released code/data and incomplete training hyperparameters. The paper is honest about the hydrogen oxalate trajectory breakdown, which partially undercuts the general extrapolation claim. Overall, the core modeling idea is sound and the empirical results are encouraging.

major comments (3)
  1. [Abstract; Conclusion; H2CO section (Figure 4)] The claim that kernel features "solve a general problem of NN-based PESs" and guarantee controlled long-range behavior is not supported by the mechanism or by the full set of experiments. Equation (9) shows that k[3,3](r,r') decays to zero as r grows, but this only ensures that the NN input vector tends to a fixed (zero) point; the subsequent softplus layers can in principle produce any output at that input, so the decay of the features does not by itself control the asymptotic energy. The paper's own hydrogen oxalate section states that when the hydrogen bond breaks and rotation about the C-C bond becomes possible, "the trajectory breaks down in such situations" because the features are not symmetrized. That is precisely an extrapolation failure for a geometry outside the training topology. The abstract and conclusion should be qualified to claim extrapolation within the same bonding topology (or for coordinates where the reference structure remains a valid anchor), and the hydrogen oxalate limitation should be prominently stated. Suggested concrete tests: systematic bond-breaking scans for more than one coordinate in a polyatomic molecule, or a demonstration that KerNN's output in the saturated-feature region is bounded by training data constraints.
  2. [H2CO section; Table S3; Figure 2] The comparison to PhysNet and other baselines for H2CO is taken from Reference 12, and it is not clear that the training/validation/test splits and data sizes are identical to those used for KerNN. Because the parity claim ("on par with PhysNet") is central, the manuscript should either retrain the baselines on the identical splits (as is done for hydrogen oxalate) or provide a clear statement that the reference errors were obtained with the same protocol. Additionally, the five repeated KerNN runs are reported only as means in Figure 2; including standard deviations or individual points would allow assessment of model variability, which matters for a model with only about 10^3 parameters.
  3. [Data Availability; Methods] The code and data are promised "upon publication," but no public repository or version is provided in the preprint, and the training hyperparameters (learning rate, batch size, number of epochs, AMSGRAD beta values) are not reported. Since the paper's contributions are empirical and rely on training a small NN, this incomplete information prevents an independent validation of the numerical results. Please provide the repository link, a data statement, and a complete hyperparameter table in the revision.
minor comments (5)
  1. [Conclusion] The sentence "the training and inference of KerNN_s is computationally more demanding than for KerNN_s" should read "than for KerNN_ns."
  2. [Hydrogen Oxalate (Methods)] In the data generation description, the temperature list "100, 300, 500, 100, 1500, 2000 K" appears to contain a typo; "100" should likely be "1000".
  3. [Table S5] The timing entries are difficult to parse because of mixed units and the bold-face markers for FORTRAN implementations; please reformat with explicit columns for each implementation and clear units.
  4. [Equation (7)] The loss function in Eq. (7) is missing the double vertical bars used for the norm in Eq. (6); please use consistent norm notation throughout.
  5. [Figure 4] The label "KerNN_s^TL" is introduced in the caption but not explained in the main text at first use; please define the transfer-learning procedure explicitly when the figure is referenced.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: KerNN's claims are benchmarked against held-out data and independent comparisons, with the one fitted constraint (KerNNs_TL) explicitly disclosed.

full rationale

The paper's central claims — small KerNN models match PhysNet accuracy, kernel features aid extrapolation, and observables are reproduced — are supported by held-out test sets, comparison to independently trained models (PhysNet, RKHS+F, FCHL) and experimental spectra, rather than by construction from the inputs. The only place experimental information enters the model is the explicitly labeled transfer-learned variant KerNNs_TL, where the experimental C-H dissociation energy is used to constrain the long-range tail; this is presented as a constrained fit, not as a prediction, so it does not constitute circularity. The extrapolation argument (Eq. 9 kernel decay leads to bounded features) is an architectural and empirical claim about generalization; whether saturation of kernel features fully controls the softplus network's output is a correctness question, not a circularity question, and the paper itself documents a breakdown in hydrogen oxalate when the hydrogen bond breaks. Self-citations (PhysNet, RKHS, earlier H2CO data) provide data or comparison methods with independent content and are not used as unverified authority to force the conclusions. Overall, no load-bearing derivation reduces to its own inputs.

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

All assumptions are standard domain assumptions for machine-learned potential energy surfaces or explicit limitations stated in the text. No new physical entities are introduced. The free parameters are the network weights plus transparently reported hyperparameters and descriptor choices.

free parameters (4)
  • NN weights and biases = 1001 to 6608 parameters depending on system (Table S1)
    Learned by minimizing the loss in Eq. 6 or Eq. 7 with AMSGRAD; the paper's central claim is that this small parameter count is sufficient.
  • Loss weights omega_F, omega_mu, omega_Q = omega_F = 10 or 1, omega_mu = 5, omega_Q = 2
    Hand-chosen hyperparameters from Table S1 that balance energy, force, and dipole errors; they affect the reported test metrics.
  • Kernel parameters [n,m] = [3,3]
    Fixed reciprocal power kernel parameters from Eq. 9, taken from prior RKHS work; they set smoothness and long-range asymptotic decay.
  • Reference distances r'_i in descriptors = Equilibrium or linear reference structures per molecule
    The descriptors in Eqs. 10 and 11 require a reference structure r'; its choice influences the descriptor values and the extrapolation behavior.
assumptions (4)
  • domain assumption Ab initio reference energies, forces, and dipole moments are accurate enough ground truth for training and evaluation.
    Used throughout the Methods; the authors note for HeH2+ near dissociation UCCSD(T) differs from FCI by up to 1.40 kcal/mol, so the reference itself is imperfect in that region.
  • domain assumption Pairwise distance kernel descriptors are sufficient to determine the energy of the molecular configurations considered.
    Descriptors in Eqs. 10 and 11 use only interatomic distances; this is standard for small rigid molecules but fails for conformational changes such as the hydrogen oxalate H-bond breaking, where the authors report trajectory breakdown.
  • domain assumption The chosen small fully connected network with softplus activations can represent the target PES within the reported accuracy.
    No formal approximation guarantee is provided for these exact architectures; the claim rests on the empirical test-set performance reported in Figures 2, 6, and 7.
  • domain assumption The quantum correction factor Q(omega)=tanh(beta hbar omega/2) is appropriate for converting classical dipole autocorrelation spectra to IR spectra.
    Used in Eq. 3 following Ref. 25; standard but an approximation for anharmonic systems.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Bigger the Better? Accurate Molecular Potential Energy Surfaces from Minimalist Neural Networks." pith.science (2026). https://pith.science/paper/LCSLNFGQ

@misc{pith2026241118121,
  author       = {Pith},
  title        = {Pith review of: The Bigger the Better? Accurate Molecular Potential Energy Surfaces from Minimalist Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LCSLNFGQ}},
  note         = {Machine review of arXiv:2411.18121}
}
read the original abstract

Atomistic simulations are a powerful tool for studying the dynamics of molecules, proteins, and materials on wide time and length scales. Their reliability and predictiveness, however, depend directly on the accuracy of the underlying potential energy surface (PES). Guided by the principle of parsimony this work introduces KerNN, a combined kernel/neural network-based approach to represent molecular PESs. Compared to state-of-the-art neural network PESs the number of learnable parameters of KerNN is significantly reduced. This speeds up training and evaluation times by several orders of magnitude while retaining high prediction accuracy. Importantly, using kernels as the features also improves the extrapolation capabilities of KerNN far beyond the coverage provided by the training data which solves a general problem of NN-based PESs. KerNN applied to spectroscopy and reaction dynamics shows excellent performance on test set statistics and observables including vibrational bands computed from classical and quantum simulations.

Figures

Figures reproduced from arXiv: 2411.18121 by the authors.

Figure 1
Figure 1. Schematic representation of A) formaldehyde, B) the two reaction channels of the [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Energy and force learning curves for the different variants of the H2CO PESs trained on CCSD(T)-F12B/aug-cc-pVTZ-F12 reference data. These are compared to Phys￾Net results taken from Reference 12. Solid and dashed lines represent MAEs and RMSEs, respectively. A total of five KerNN models were trained for each value of NTrain on different splits of the data and only the mean out-of-sample errors are shown. KerNNns an… view at source ↗
Figure 3
Figure 3. , illustrates that the atomic MAE(F) are larger for the structure using KerNNns . Also, KerNNns fails to predict fully symmetric forces (with atomic MAE(F) of 0.99 and 0.80 kcal/mol for the two H-atoms) whereas for KerNNs the atomic MAE(F) is identical on both H-atoms (0.39 kcal/mol) [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: A: The extrapolation capabilities of the ML-PES are assessed on a data set con [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Infrared spectra derived from finite-T MD simulations of H2CO. The experimental fundamentals23 are the grey Gaussians. The computed spectra were averaged over 100 independent trajectories, each 200 ps in length, using ∆t = 0.2 fs. The MD simulations, carried out using …
Figure 6
Figure 6. Figure 6: Out-of-sample errors for the HeH+ 2 KerNNns (A, cyan) and KerNNs (B, olive) PESs (with Dns and Ds as descriptors, respectively) trained on UCCSD(T)/aug-cc-pV5Z level reference data. The test set contains 4709 randomly chosen structures. Most energies are predicted with…
Figure 7
Figure 7. Figure 7: Out of sample errors on a test set containing 2000 hydrogen oxalate structures for [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: The framework modes below 2000 cm−1 were all captured rather accurately com￾pared with experiment. It should be noted that the PESs are based on the MP2 level of theory and that the experiments were carried out using the H2-messenger technique which, strictly speaking,…
Figure 8
Figure 8. Figure 8: Infrared spectra of hydrogen oxalate. The computed spectra (top two traces) are [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Dynamics of Protonated Oxalate from Machine-Learned Simulations and Experiment: Infrared Signatures, Proton Transfer Dynamics and Tunneling Splittings

    physics.chem-ph 2025-08 unverdicted novelty 6.0 of 10

    Oxalate spectra and proton dynamics were reportedly simulated with a CCSD(T)-quality machine-learned PES predicting a 35 cm^-1 tunneling splitting, but the supplied text is a different paper.

  2. Towards Large-Scale Condensed Phase Simulations using Machine Learned Energy Functions

    physics.chem-ph 2025-06 conditional novelty 5.0 of 10

    A machine-learned water model combining a CCSD(T)-quality monomer neural network, flexible distributed charges, and cluster-fitted Lennard-Jones terms reproduces many bulk liquid properties in multi-nanosecond simulat...

Reference graph

Works this paper leans on

69 extracted references · 53 canonical work pages · cited by 2 Pith papers

  1. [1]

    F.; Berendsen, H

    Van Gunsteren, W. F.; Berendsen, H. J. Computer simulation of molecular dynamics: methodology, applications, and perspectives in chemistry. Angew. Chem., Int. Ed. Engl. 1990, 29, 992--1023

    work page 1990

  2. [2]

    Karplus, M.; Petsko, G. A. Molecular dynamics simulations in biology. Nature 1990, 347, 631--639

    work page 1990

  3. [3]

    D.; McCammon, J

    Durrant, J. D.; McCammon, J. A. Molecular dynamics simulations and drug discovery. BMC Biology 2011, 9, 1--9

    work page 2011

  4. [4]

    Vanommeslaeghe, K.; Hatcher, E.; Acharya, C.; Kundu, S.; Zhong, S.; Shim, J.; Darian, E.; Guvench, O.; Lopes, P.; Vorobyov, I. et al. CHARMM general force field: A force field for drug-like molecules compatible with the CHARMM all-atom additive biological force fields. J. Comput. Chem. 2010, 31, 671--690

    work page 2010

  5. [5]

    L.; Blondel, A.; Boittier, E

    Hwang, W.; Austin, S. L.; Blondel, A.; Boittier, E. D.; Boresch, S.; Buck, M.; Buckner, J.; Caflisch, A.; Chang, H.-T.; Cheng, X. et al. CHARMM at 45: Enhancements in accessibility, functionality, and speed. J. Phys. Chem. B 2024, 128, 9976--10042

    work page 2024

  6. [6]

    utt, K. T.; Tkatchenko, A.; M\

    Unke, O. T.; Chmiela, S.; Sauceda, H. E.; Gastegger, M.; Poltavsky, I.; Sch\"utt, K. T.; Tkatchenko, A.; M\"uller, K.-R. Machine learning force fields. Chem. Rev. 2021, 121, 10142--10186

    2021

  7. [7]

    a ser, S.; Vazquez-Salazar, L. I.; Meuwly, M.; T \

    K \"a ser, S.; Vazquez-Salazar, L. I.; Meuwly, M.; T \"o pfer, K. Neural network potentials for chemistry: concepts, applications and prospects. Digital Discovery 2023, 2, 28--58

    work page 2023

  8. [8]

    L.; Bowman, J

    Qu, C.; Yu, Q.; Van Hoozen Jr, B. L.; Bowman, J. M.; Vargas-Hern \'a ndez, R. A. Assessing Gaussian process regression and permutationally invariant polynomial approaches to represent high-dimensional potential energy surfaces. J. Chem. Theory Comput. 2018, 14, 3381--3396

    work page 2018

Show all 69 references
  1. [9]

    S.; Wieder , M.; Xu , Y.; Zhu , T.; Zhang , J

    Wang , Y.; Takaba , K.; Chen , M. S.; Wieder , M.; Xu , Y.; Zhu , T.; Zhang , J. Z. H.; Nagle , A.; Yu , K.; Wang , X. et al. On the design space between molecular mechanics and machine learning force fields . arXiv e-prints 2024, arXiv:2409.01931

    2024 arXiv

  2. [10]

    Computational invariant theory; Springer, 2015

    Derksen, H.; Kemper, G. Computational invariant theory; Springer, 2015

    2015

  3. [11]

    G.; Pearlmutter, B

    Baydin, A. G.; Pearlmutter, B. A.; Radul, A. A.; Siskind, J. M. Automatic differentiation in machine learning: a survey. J. Mach. Learn. Res. 2017, 18, 5595--5637

    2017

  4. [12]

    S.; von Lilienfeld, O

    K \"a ser, S.; Koner, D.; Christensen, A. S.; von Lilienfeld, O. A.; Meuwly, M. Machine Learning Models of Vibrating H _2 CO: Comparing Reproducing Kernels, FCHL, and PhysNet. J. Phys. Chem. A 2020, 124, 8853--8865

    2020

  5. [13]

    A numerical study on learning curves in stochastic multilayer feedforward networks

    M \"u ller, K.-R.; Finke, M.; Murata, N.; Schulten, K.; Amari, S.-I. A numerical study on learning curves in stochastic multilayer feedforward networks. Neural Comput. 1996, 8, 1085--1106

    1996

  6. [14]

    Huang, B.; Von Lilienfeld, O. A. Communication: Understanding molecular representations in machine learning: The role of uniqueness and target similarity. J. Chem. Phys. 2016, 145, 161102

    2016

  7. [15]

    S.; Von Lilienfeld, O

    Christensen, A. S.; Von Lilienfeld, O. A. On the role of gradients for machine learning of molecular energies and forces. Mach. Learn.: Sci. Technol. 2020, 1, 045018

    2020

  8. [16]

    T.; Meuwly, M

    Unke, O. T.; Meuwly, M. PhysNet: A neural network for predicting energies, forces, dipole moments, and partial charges. J. Chem. Theory Comput. 2019, 15, 3678--3693

    2019

  9. [17]

    Permutationally invariant, reproducing kernel-based potential energy surfaces for polyatomic molecules: From formaldehyde to acetone

    Koner, D.; Meuwly, M. Permutationally invariant, reproducing kernel-based potential energy surfaces for polyatomic molecules: From formaldehyde to acetone. J. Chem. Theory Comput. 2020, 16, 5474--5484

    2020

  10. [18]

    A.; Christensen, A

    Faber, F. A.; Christensen, A. S.; Huang, B.; Von Lilienfeld, O. A. Alchemical and structural distribution based representation for universal quantum machine learning. J. Chem. Phys. 2018, 148, 241717

    2018

  11. [19]

    J.; Chmiela, S.; Cui, M.; Frank, J

    Poltavsky, I.; Charkin-Gorbulin, A.; Puleva, M.; Cordeiro Fonseca, G.; Batatia, I.; Browning, N. J.; Chmiela, S.; Cui, M.; Frank, J. T.; Heinen, S. et al. Crash Testing Machine Learning Force Fields for Molecules, Materials, and Interfaces: Model Analysis in the TEA Challenge ...

    2023 doi

  12. [20]

    J.; Chmiela, S.; Cui, M.; Frank, J

    Poltavsky, I.; Puleva, M.; Charkin-Gorbulin, A.; Cordeiro Fonseca, G.; Batatia, I.; Browning, N. J.; Chmiela, S.; Cui, M.; Frank, J. T.; Heinen, S. et al. Crash Testing Machine Learning Force Fields for Molecules, Materials, and Interfaces: Molecular Dynamics in the TEA Challe...

    2023 doi

  13. [21]

    F.; Moore, C

    Chuang, M.-C.; Foltz, M. F.; Moore, C. B. T_1 barrier height, S_1--T_1 intersystem crossing rate, and S_0 radical dissociation threshold for H _2 CO, D _2 CO, and HDCO. J. Chem. Phys. 1987, 87, 3855--3864

    1987

  14. [22]

    Dawes, R.; Ndengu \'e , S. A. Single- and multireference electronic structure calculations for constructing potential energy surfaces. Intern. Rev. Phys. Chem. 2016, 35, 441--478

    2016

  15. [23]

    C.; Nelson Jr, D

    Herndon, S. C.; Nelson Jr, D. D.; Li, Y.; Zahniser, M. S. Determination of line strengths for selected transitions in the v_2 band relative to the v_1 and v_5 bands of H2CO. J. Quant. Spectrosc. Radiat. Transf. 2005, 90, 207--216

    2005

  16. [24]

    D.; Meuwly, M

    T \"o pfer, K.; Koner, D.; Erramilli, S.; Ziegler, L. D.; Meuwly, M. Molecular-level understanding of the rovibrational spectra of N _2 O in gaseous, supercritical, and liquid SF _6 and Xe. J. Chem. Phys. 2023, 158, 144302

    2023

  17. [25]

    Quantum corrections to classical time-correlation functions: Hydrogen bonding and anharmonic floppy modes

    Kumar, P.; Marx, D. Quantum corrections to classical time-correlation functions: Hydrogen bonding and anharmonic floppy modes. J. Chem. Phys 2004, 121, 9

    2004

  18. [26]

    H.; Mortensen, J

    Larsen, A. H.; Mortensen, J. J.; Blomqvist, J.; Castelli, I. E.; Christensen, R.; Du ak, M.; Friis, J.; Groves, M. N.; Hammer, B.; Hargus, C. et al. The atomic simulation environment -- a Python library for working with atoms. J. Phys. Condens. Matter 2017, 29, 273002

    2017

  19. [27]

    Bibliographic review and new measurements of the infrared band strengths of pure molecules at 25 K: H _2 O, CO _2 , CO, CH _4 , NH _3 , CH _3 OH, HCOOH and H _2 CO

    Bouilloud, M.; Fray, N.; B \'e nilan, Y.; Cottin, H.; Gazeau, M.-C.; Jolly, A. Bibliographic review and new measurements of the infrared band strengths of pure molecules at 25 K: H _2 O, CO _2 , CO, CH _4 , NH _3 , CH _3 OH, HCOOH and H _2 CO. Mon. Not. R. Astron. Soc. 2015, 4...

    2015

  20. [28]

    T.; Meuwly, M

    Unke, O. T.; Meuwly, M. Toolkit for the construction of reproducing kernel-based representations of data: Application to multidimensional potential energy surfaces. J. Chem. Inf. and Mod. 2017, 57, 1923--1931

    2017

  21. [29]

    A.; Barletta, P.; Harris, G

    Tennyson, J.; Kostin, M. A.; Barletta, P.; Harris, G. J.; Polyansky, O. L.; Ramanlal, J.; Zobov, N. F. DVR3D: a program suite for the calculation of rotation--vibration spectra of triatomic molecules. Comput. Phys. Commun. 2004, 163, 85--116

    2004

  22. [30]

    Koner, D.; Veliz, J. C. S. V.; van der Avoird, A.; Meuwly, M. Near dissociation states for H _2^+ --He on MRCI and FCI potential energy surfaces. Phys. Chem. Chem. Phys. 2019, 21, 24976--24983

    2019

  23. [31]

    Asvany, O.; Schlemmer, S.; van der Avoird, A.; Szidarovszky, T.; Cs \'a sz \'a r, A. G. Vibrational spectroscopy of H _2 He ^+ and D _2 He ^+ . J. Mol. Spectrosc. 2021, 377, 111423

    2021

  24. [32]

    P.; Reich, D

    Margulis, B.; Horn, K. P.; Reich, D. M.; Upadhyay, M.; Kahn, N.; Christianen, A.; van der Avoird, A.; Groenenboom, G. C.; Meuwly, M.; Koch, C. P. et al. Tomography of Feshbach resonance states. Science 2023, 380, 77--81

    2023

  25. [33]

    P.; Vazquez-Salazar, L

    Horn, K. P.; Vazquez-Salazar, L. I.; Koch, C. P.; Meuwly, M. Improving potential energy surfaces using measured Feshbach resonance states. Sci. Adv. 2024, 10, eadi6462

    2024

  26. [34]

    P.; Upadhyay, M.; Margulis, B.; Reich, D

    Horn, K. P.; Upadhyay, M.; Margulis, B.; Reich, D. M.; Narevicius, E.; Meuwly, M.; Koch, C. P. Feshbach resonances in cold collisions as a benchmark for state of the art ab initio theory. arXiv preprint arXiv:2408.13197 2024,

    2024 arXiv

  27. [35]

    T.; DeBlase, A

    Wolke, C. T.; DeBlase, A. F.; Leavitt, C. M.; McCoy, A. B.; Johnson, M. A. Diffuse vibrational signature of a single proton embedded in the oxalate scaffold, HO _2 CCO _2^- . J. Phys. Chem. A 2015, 119, 13018--13024

    2015

  28. [36]

    Vibrational spectroscopy and proton transfer dynamics in protonated oxalate

    Xu, Z.-H.; Meuwly, M. Vibrational spectroscopy and proton transfer dynamics in protonated oxalate. J. Phys. Chem. A 2017, 121, 5389--5398

    2017

  29. [37]

    T.; Meuwly, M

    K \"a ser, S.; Unke, O. T.; Meuwly, M. Reactive dynamics and spectroscopy of hydrogen transfer from neural network-based reactive potential energy surfaces. New J. Phys. 2020, 22, 055002

    2020

  30. [38]

    Blumer, A.; Ehrenfeucht, A.; Haussler, D.; Warmuth, M. K. Occam's razor. Inf. Process. Lett. 1987, 24, 377--380

    1987

  31. [39]

    The role of Occam's razor in knowledge discovery

    Domingos, P. The role of Occam's razor in knowledge discovery. Data Min. Knowl. Discov. 1999, 3, 409--425

    1999

  32. [40]

    A.; Bach , N.; Bahree , A.; Bakhtiari , A.; Bao , J.; Behl , H

    Abdin , M.; Aneja , J.; Awadalla , H.; Awadallah , A.; Awan , A. A.; Bach , N.; Bahree , A.; Bakhtiari , A.; Bao , J.; Behl , H. et al. Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone . arXiv e-prints 2024, arXiv:2404.14219

    2024 arXiv

  33. [41]

    Tan , M.; Le , Q. V. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks . arXiv e-prints 2019, arXiv:1905.11946

    2019 arXiv

  34. [42]

    F.; Pozdnyakov , S

    Langer , M. F.; Pozdnyakov , S. N.; Ceriotti , M. Probing the effects of broken symmetries in machine learning . arXiv e-prints 2024, arXiv:2406.17747

    2024 arXiv

  35. [43]

    C.; Bemish, R

    Koner, D.; San Vicente Veliz, J. C.; Bemish, R. J.; Meuwly, M. Accurate reproducing kernel-based potential energy surfaces for the triplet ground states of N _2 O and dynamics for the N + NO O + N _2 and N _2 + O 2N + O reactions. Phys. Chem. Chem. Phys. 2020, 22, 18488--18498

    2020

  36. [44]

    J.; Meuwly, M

    Koner, D.; Bemish, R. J.; Meuwly, M. The C( ^3 P) + NO(X ^2 ) O( ^3 P) + CN(X ^2 ^+ ), N( ^2 D)/N( ^4 S) + CO(X ^1 ^+ ) reaction: Rates, branching ratios, and final states from 15 K to 20 000 K . J. Chem. Phys. 2018 , 149 , 094305

    2018

  37. [45]

    S.; Isayev, O.; Roitberg, A

    Smith, J. S.; Isayev, O.; Roitberg, A. E. ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost. Chem. Sci. 2017, 8, 3192--3203

    2017

  38. [46]

    u tz , M.; Celani, P.; Gy\

    Werner, H.-J.; Knowles, P. J.; Knizia, G.; Manby, F. R.; Sch\" u tz , M.; Celani, P.; Gy\"orffy, W.; Kats, D.; Korona, T.; Lindh, R. et al. MOLPRO, version 2019, a package of ab initio programs. 2019

    2019

  39. [47]

    GFN2-xTB—An accurate and broadly parametrized self-consistent tight-binding quantum chemical method with multipole electrostatics and density-dependent dispersion contributions

    Bannwarth, C.; Ehlert, S.; Grimme, S. GFN2-xTB—An accurate and broadly parametrized self-consistent tight-binding quantum chemical method with multipole electrostatics and density-dependent dispersion contributions. J. Chem. Theory Comput. 2019, 15, 1652--1671

    2019

  40. [48]

    Introduction to the diffusion Monte Carlo method

    Kosztin, I.; Faber, B.; Schulten, K. Introduction to the diffusion Monte Carlo method. Am. J. Phys. 1996, 64, 633--644

    1996

  41. [49]

    Transfer learned potential energy surfaces: accurate anharmonic vibrational dynamics and dissociation energies for the formic acid monomer and dimer

    K \"a ser, S.; Meuwly, M. Transfer learned potential energy surfaces: accurate anharmonic vibrational dynamics and dissociation energies for the formic acid monomer and dimer. Phys. Chem. Chem. Phys. 2022, 24, 5269--5281

    2022

  42. [50]

    P.; Ba, J

    Kingma, D. P.; Ba, J. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 2014,

    2014 arXiv

  43. [51]

    Neural Networks: Tricks of the trade; Springer, 2002; pp 55--69

    Prechelt, L. Neural Networks: Tricks of the trade; Springer, 2002; pp 55--69

    2002

  44. [52]

    Raghunathan, S.; Priyakumar, U. D. Molecular representations for machine learning applications in chemistry. Int. J. Quantum Chem. 2022, 122, e26870

    2022

  45. [53]

    W.; Behler, J

    Kocer, E.; Ko, T. W.; Behler, J. Neural network potentials: A concise overview of methods. Ann. Rev. Phys. Chem. 2022, 73, 163--186

    2022

  46. [54]

    Generalized neural-network representation of high-dimensional potential-energy surfaces

    Behler, J.; Parrinello, M. Generalized neural-network representation of high-dimensional potential-energy surfaces. Phys. Rev. Lett. 2007, 98, 146401

    2007

  47. [55]

    Zou, S.; Bowman, J. M. A new ab initio potential energy surface describing acetylene/vinylidene isomerization. Chem. Phys. Lett. 2003, 368, 421--424

    2003

  48. [56]

    V.; Acioli, P

    Prudente, F. V.; Acioli, P. H.; Neto, J. S. The fitting of potential energy surfaces using neural networks: Application to the study of vibrational levels of H 3+. J. Chem. Phys. 1998, 109, 8801--8808

    1998

  49. [57]

    T.; Le, H

    Nguyen, H. T.; Le, H. M. Modified feed-forward neural network structures and combined-function-derivative approximations incorporating exchange symmetry for potential energy surface fitting. J. Phys. Chem. A 2012, 116, 4629--4638

    2012

  50. [58]

    Permutation invariant polynomial neural network approach to fitting potential energy surfaces

    Jiang, B.; Guo, H. Permutation invariant polynomial neural network approach to fitting potential energy surfaces. J. Chem. Phys. 2013, 139, 054112

    2013

  51. [59]

    Shao, K.; Chen, J.; Zhao, Z.; Zhang, D. H. Communication: Fitting potential energy surfaces with fundamental invariant neural network. J. Chem. Phys. 2016, 145, 071101

    2016

  52. [60]

    J.; Bowman, J

    Braams, B. J.; Bowman, J. M. Permutationally invariant potential energy surfaces in high dimensionality. Int. Rev. Phys. Chem. 2009, 28, 577--606

    2009

  53. [61]

    Xie, Z.; Bowman, J. M. Permutationally invariant polynomial basis for molecular energy surface fitting via monomial symmetrization. J. Chem. Theory Comput. 2010, 6, 26--34

    2010

  54. [62]

    M.; Czako, G.; Fu, B

    Bowman, J. M.; Czako, G.; Fu, B. High-dimensional ab initio potential energy surfaces for reaction dynamics calculations. Phys. Chem. Chem. Phys. 2011, 13, 8094--8111

    2011

  55. [63]

    Theory of reproducing kernels

    Aronszajn, N. Theory of reproducing kernels. Trans. Am. Math. Soc. 1950, 68, 337--404

    1950

  56. [64]

    A general method for constructing multidimensional molecular potential energy surfaces from ab initio calculations

    Ho, T.-S.; Rabitz, H. A general method for constructing multidimensional molecular potential energy surfaces from ab initio calculations . J. Chem. Phys. 1996, 104, 2584

    1996

  57. [65]

    C.; Koner, D.; Schwilk, M.; Bemish, R

    San Vicente Veliz, J. C.; Koner, D.; Schwilk, M.; Bemish, R. J.; Meuwly, M. The N (^4S) + O _2(X^3 ^-_g) O (^3P) + NO (X^2 ) Reaction: Thermal and Vibrational Relaxation Rates for the ^ 2 A ' , ^ 4 A ' and ^ 2 A '' States . Phys. Chem. Chem. Phys. 2020, 22, 3927--3939

    2020

  58. [66]

    Formation and Stabilization of Ground and Excited-State Singlet O _2 upon Recombination of ^3 P Oxygen on Amorphous Solid Water

    Pezzella, M.; Koner, D.; Meuwly, M. Formation and Stabilization of Ground and Excited-State Singlet O _2 upon Recombination of ^3 P Oxygen on Amorphous Solid Water . J. Phys. Chem. Lett. 2020, 11, 2171--2176

    2020

  59. [67]

    Sold \'a n, P.; Hutson, J. M. On the long-range and short-range behavior of potentials from reproducing kernel Hilbert space interpolation. J. Chem. Phys. 2000, 112, 4415--4416

    2000

  60. [68]

    King, S. A. Minimal generating sets of non-modular invariant rings of finite groups. J. Symbolic Comput. 2013, 48, 101--109

    2013

  61. [69]

    error floor

    Decker, W.; Greuel, G.-M.; Pfister, G.; Sch\"onemann, H. Singular 4-3-0 --- A computer algebra system for polynomial computations. http://www.singular.uni-kl.de, 2022 mcitethebibliography main.tex0000664000000000000000000023122114721352706011234 0ustar rootroot [journal=jpcafh...

    2022

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.