Pith. sign in

REVIEW 3 major objections 5 minor 40 references

Neuro-Symbolic Operator for Interpretable and Generalizable Characterization of Complex Piezoelectric Systems

T0 review · 3 major / 5 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 This paper proposes a neuro-symbolic operator (NSO) that distills a Fourier neural operator into a sparse, interpretable ODE governing piezoelectric hysteresis, and shows this discovered ODE predicts displacement for voltage fields…

desk verdict The noise-robustness experiment is clean, but the butterfly results are unreproducible as written and the OOD comparison is structurally unfair. read the letter →

arxiv 2505.24578 v1 pith:GR4HSJVL submitted 2025-05-30 cs.LG

Abhishek Chandra , Taniya Kapoor , Mitrofan Curti , Koen Tiels , Elena A. Lomonova This is my paper · ORCID
classification cs.LG MSC 68T07
keywords neuro-symbolicoperatorneuraloperatorspiezoelectrichysteresissparsemodeldiscoveryFourierout-of-distributiongeneralizationbutterflyinterpretablemodeling
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 proposes a two-stage neuro-symbolic operator (NSO) for piezoelectric hysteresis. It first trains a Fourier neural operator mapping voltage fields to displacement profiles, then uses sequential threshold least-squares regression on the operator's predictions to discover a sparse, interpretable ordinary differential equation that governs the hysteresis. The central claim is that this white-box ODE not only matches the training data but predicts displacement profiles for voltage fields drawn from different stochastic-process kernels than anything seen in training, including butterfly-shaped hysteresis loops. The motivation is that standard neural operators are black boxes that fail outside their training input distribution, whereas an explicit ODE can be inspected, simulated, and applied to new voltage inputs.

What carries the argument

The load-bearing mechanism is the two-stage distillation: Stage I uses a Fourier neural operator with $K$ spectral convolution layers to learn a smooth surrogate $\hat{P}=N_\theta: \mathcal{V}\to\mathcal{D}$, which denoises and upsamples the data; Stage II assembles a library $\Theta(D,V,f(V))$ of candidate terms built from voltage, displacement, derivatives, absolute values, and degree-2 polynomial combinations, then solves the sparse regression $\dot{D}=\Theta(D,V,f(V))\xi$ via sequential threshold least squares. The resulting sparse coefficient vector $\xi^*$ selects the few terms that define the discovered ODE. What makes the argument work is that the neural operator's predictions are clean enough for derivative estimation, so the sparse regression can pick out the true structure from a library that already contains it.

What would settle it

Generate synthetic voltage-displacement data from a hysteresis ODE whose library is deliberately missing a term, e.g. $\dot{d}=0.4|\dot{v}|v-0.85|\dot{v}|d+0.2\dot{v}+0.5\sqrt{|\dot{v}|}$, run NSO with the paper's library, and check whether the recovered ODE either omits the $\sqrt{|\dot{v}|}$ term or fails to predict a held-out RBF voltage field; either outcome would refute the claimed generalizability.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a neural operator's learned mapping can be distilled into a parsimonious differential model without losing predictive accuracy. Concretely, training an FNO on Sine-kernel voltage inputs and then applying STLSQ to the FNO's predicted displacement profiles recovers equations such as $\dot{d} = 0.39|\dot{v}|v - 0.83|\dot{v}|d + 0.2\dot{v}$ for a standard hysteresis law and two-equation systems with a latent variable $y$ for butterfly-shaped hysteresis. These recovered ODEs, when solved forward, give accurate displacement predictions for RBF and Matern voltage fields, whereas the neural operators alone degrade by an order of magnitude. The paper takes this as evidence that the discovered symbolic model, not the neural network, carries the generalizable content of the hysteresis operator.

Load-bearing premise

The whole claim rests on the candidate library containing the exact functional forms of the true hysteresis law, and on any latent internal variable $y$ being present or recoverable in that library; if a real system has terms the library does not include, the discovered ODE will not be the true law and out-of-distribution generalization will fail.

Editorial extensions

If this is right

  • NSO's discovered ODEs generalize to RBF and Matern voltage fields even though training used only Sine fields, with errors one to two orders of magnitude lower than standalone neural operators.
  • The white-box ODE form lets engineers inspect which terms govern the hysteresis instead of treating the model as a black box.
  • Because Stage I denoises and upsamples, NSO recovers nearly exact coefficients even with 20% Gaussian noise and with only 20 time samples per function.
  • The framework extends from simple hysteresis loops to butterfly-shaped hysteresis by including a latent variable $y$ in the discovered ODE system.
  • The sparsity threshold $\lambda$ controls a trade-off between accuracy and interpretability, with the paper reporting that $\lambda=0.01$ gives a good balance.

Reading between the lines

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

  • The method's core assumption is that the hysteresis law is sparse in a human-chosen library; if that is accepted, NSO is effectively an equation-discovery pipeline that uses a neural operator as a data-cleaning front end.
  • A natural extension is to apply NSO to measured experimental voltage-displacement data where no closed-form law is known and validate the discovered ODE on a completely different voltage class, something the paper demonstrates only on synthetic ODE-generated data.
  • The butterfly experiments suggest the latent state $y$ is identifiable from trajectories, but the paper does not provide a recovery procedure; whether this generalizes to hysteresis with multiple or unobservable internal states is open.
  • If the discovered ODE equals the true law, then NSO would make neural operators unnecessary after training: deployment could use the ODE directly, a consequence the paper leaves implicit.
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 proposes a neuro-symbolic operator (NSO) framework for hysteresis modeling. Stage I trains a Fourier neural operator on voltage-displacement pairs sampled from one input family (Sine kernel). Stage II uses the FNO predictions on Sine test inputs to build a candidate library of polynomial and absolute-value terms and runs sequential threshold least-squares to extract a sparse ODE for the displacement. The identified ODE is then simulated on held-out voltage fields from RBF and Matérn kernels. The paper reports seven experiments: two first-order hysteresis laws, two butterfly-shaped two-state hysteresis laws, noise and low-fidelity ablations, and a sparsity-threshold ablation.

Significance. If the results were fully supported, NSO would be a useful bridge between neural operators and symbolic model discovery: Experiments 1-2 recover coefficients close to the generating equations, and the noise/fidelity ablations show a practical advantage of using an operator surrogate as a denoiser before sparse regression. However, the butterfly experiments and the model-agnosticism claim depend on an undeclared latent variable and on a library that is constructed from the true equation terms. The central generalization claim is therefore not established in the current version.

major comments (3)
  1. [§IV-B, §V-A, §V-D] Experiments 3 and 4 report discovered ODEs in the state (d, y), for example ˙d = 2|v̇|vy − 4.70|v̇|dy + 3v̇y and ˙y = |v̇|v − 2.35|v̇|y + 1.5v̇, but the candidate library Θ(D,V,f(V)) defined in Section IV-B and the STLSQ basis described in Section V-A contain only voltage fields, their derivatives, absolute valued fields and their derivatives, and displacement profiles with polynomial features up to degree two. Terms such as v y, d y, and |v̇| y cannot be formed from these basis functions, y is not an output of the FNO, and Algorithm 1 contains no latent-state reconstruction step. The butterfly results therefore cannot be produced by the described method. If y is available from the synthetic data generator, that use must be disclosed and a deployment-relevant procedure for obtaining y must be supplied; as written, the claim that NSO discovers these two-state models is unsupported.
  2. [§VI and §V-A] The discussion in Section VI claims that NSO 'does not require explicitly including candidate terms governing the dynamics in the library' and is model-agnostic, but this is contradicted by the experimental setup. The candidate library is built from exactly the voltage, derivative, absolute-value, and displacement terms appearing in the generating equations, and the recovered ODEs in Experiments 1-2 are precisely the monomials contained in the library. The library composition is a free design choice, and no misspecification experiment is reported. The paper should either demonstrate discovery with an incomplete or perturbed library or substantially weaken the model-agnosticism claim.
  3. [Table I] Table I contains an internally inconsistent DON column. For each experiment the DON RMSE is identical across Sine, RBF, and Matern rows (8.63e-02 in Exp 1, 6.14e-02 in Exp 2, 4.02e-02 in Exp 3, and 3.11e-01 in Exp 4), while the corresponding R and MAE values differ across those rows. Such a pattern cannot arise from independent evaluations on different test sets. The table needs to be corrected or an explanation provided before the claim that all baseline operators fail out-of-distribution can be accepted.
minor comments (5)
  1. [§V-A] The description of the second stage should state explicitly that NSO's RBF and Matern predictions are obtained by simulating the discovered ODE with the test voltage fields, not by evaluating the FNO on those fields; Table I and Table IV as presented make it difficult to distinguish the two predictors.
  2. [§V-G/Table III] Experiment 6 is said to use the Experiment 4 data with the Matern32 kernel, but Table III labels the third kernel as Matern52; unify the notation.
  3. [§V-G] In the last paragraph of Section V-G, 'NNO' should be 'NSO'.
  4. [§IV-B] The notation f(V) is never defined; the paper should list the exact basis functions included in Θ, including whether |d| and |y| are included, since the reported Exp 2 and Exp 4 equations require terms such as v̇|d|.
  5. [§V-A] The sentence about Stage II using 500 functions based on FNO predictions is ambiguous; clarify whether these 500 are disjoint from the 1000 test functions used for the other baselines and whether the same functions are also used in the error metrics of Table I.

Circularity Check

3 steps flagged · score 6.0 of 10

Exp 3–4 are circular as presented: the 'identified' butterfly ODEs contain a latent y absent from the Stage-II library Θ(D,V,f(V)) and from FNO outputs, so only the [24] generator can supply them, making the RBF/Matern errors the generator tested on its own data. Exp 1–2 are a genuine, though library-tailored, SINDy-on-FNO chain; their Sine-kernel metrics are in-sample fits.

  1. other [Section V-D (Exp. 3) and Section V-E (Exp. 4); library specification in Section IV-B and Section V-A; generator cited as [24]]
    "The dynamics governing the butterfly-shaped hysteresis are taken to be a system of equation [24]: ˙d = 2| ˙v|vy−4.70| ˙v|dy+ 3 ˙vy; ˙y = | ˙v|v − 2.35| ˙v|y + 1.5 ˙v. Here y denotes a latent variable. The model identified by NSO is: ˙d = 1.98| ˙v|vy − 4.67| ˙v|dy + 2.98 ˙v y; ˙y = 0.99| ˙v|v − 2.32| ˙v|y + 1.49 ˙v, which preserves the original structure of a white-box ODE."

    Stage II's search space is Eq. (3), ˙D = Θ(D, V, f(V))ξ, and Section V-A fixes Θ as degree-2 products of 'voltage fields, their derivatives, absolute valued fields and their derivatives, and the displacement profiles' — no latent variable appears. Stage I outputs displacement only (bdi = bQ σ(Qu(K))), and no y-reconstruction procedure is given anywhere. Therefore the reported Exp-3/4 models, coupled ODEs in d and y with terms vy, dy, v̇y, |v̇|y, cannot be outputs of Algorithm 1. Term-for-term they equal the V-D/V-E generators (cited to the authors' own [24]) with coefficients scaled by ~0.99 (2→1.98, 4.70→4.67, 3→2.98, 1→0.99, 2.35→2.32, 1.5→1.49).

  2. fitted input called prediction [Section V-A (Experiment Setting, second-stage data) and Table I (Sine columns)]
    "As the second stage of NSO utilizes the testing data of FNO for the Sine kernel case, it is trained and tested on 500 functions based on the predictions of FNO on the Sine kernel."

    STLSQ fits ˙D = Θξ on FNO's predictions for those 500 Sine-kernel functions, and the same 500 functions are used for the reported Sine-kernel evaluation: the ODE is 'trained and tested' on one set. Table I's NSO Sine columns (e.g., Exp 1 Sine R=8.11e-03) are therefore in-sample residuals of the Stage-II fit, presented under the heading of prediction accuracy. The RBF/Matern columns are genuine extrapolation of the fitted ODE, so this does not invalidate the generalization claim by itself, but the Sine columns conflate fit with prediction.

1 more flagged steps
  1. ansatz smuggled in via citation [Section VI (Discussion, 'model agnosticism') vs. Section IV-B (library construction), Section V-A (library terms), and [24]]
    "NSO does not require explicitly including candidate terms governing the dynamics in the library and discovers the dynamics without embedded bias toward the data source."

    This agnosticism claim is contradicted by the method's own equations: Eq. (3) restricts the discovered model to a sparse combination of library columns, and Section IV-B builds that library from 'known hysteresis models such as Bouc–Wen and generic polynomial and trigonometric terms,' with Section V-A enumerating exactly the monomial families used as the Exp-1–4 generators (all from the same authors' prior work [24]). The recovered Exp-1/2 equations (0.39|v̇|v − 0.83|v̇|d + 0.2v̇; 4.91v̇ − 0.25|v̇|d − 0.48v̇|d|) are precisely the library's degree-2 products, so the 'discovery' is selection inside an ansatz built from the generating model families.

full rationale

The claimed derivation chain is: (Stage I) train FNO on Sine-kernel voltage–displacement pairs; (Stage II) use FNO's Sine predictions to fit ˙D = Θ(D,V,f(V))ξ by STLSQ; then simulate the fitted ODE on RBF/Matern fields and report the error as NSO's out-of-distribution prediction (Section V-A: 'The resulting sparse model is then embedded into a first-order ODE and solved using odeint'). The FNO itself is never evaluated OOD — Table I shows FNO failing on RBF/Matern while NSO's columns come entirely from ODE simulation. For Experiments 1–2 this chain is internally coherent: the ODE is fitted only on Sine-derived data and extrapolated to new fields, which is a genuine test of the discovered law, and the recovered coefficients (0.39/0.4, 0.83/0.85, 4.91/5, 0.48/0.5) track the generators. The circularity is concentrated in the butterfly showcase. Section IV-B defines the library over predicted displacements D, voltage V, and f(V), and Section V-A restricts it to 'voltage fields, their derivatives, absolute valued fields and their derivatives, and the displacement profiles' with degree-2 polynomials; a library over D and V cannot contain vy, dy, |v̇|y, or v̇y. y is never measured, never output by FNO, and never reconstructed, yet the reported Exp-3/4 models are two-state ODEs in d and y whose terms and coefficients match the generators 'taken to be a system of equation [24]' — the same authors' prior work. By the paper's own equations these models cannot be outputs of Algorithm 1; the reading consistent with the text is that NSO's reported butterfly predictions are simulations of the [24] generator, i.e., the input model predicting data generated from itself (coefficient match within 1–2%). The Discussion's agnosticism claim is also contradicted by Eq. (3), which restricts solutions to library columns chosen from known hysteresis model families, and by the fact that the Exp-1/2 generators are exactly the library's monomials. The Sine-kernel NSO metrics are in-sample fit residuals because STLSQ is 'trained and tested' on the same 500 Sine functions. Overall: Exp 1–2 and the noise/fidelity ablations contain genuine independent content, so the score is 6 rather than higher, but the headline butterfly result reduces, as described, to the authors' prior generator.

Assumptions & free parameters 3 free parameters · 3 assumptions · 1 invented entities

The framework's central claim rests on a hand-built library, a recoverable latent state, and the assumption that the true physical law is a sparse ODE in that library; all are guaranteed by the synthetic data and the authors' prior modeling choices, not by independent evidence.

free parameters (3)
  • STLSQ sparsity threshold lambda = 0.01 default, 0.1 and 0.001 in ablation
    Chosen by hand in Section V-F and VI; controls the number of terms in the discovered ODE and directly affects predictive accuracy vs interpretability trade-off.
  • FNO frequency modes nm and hidden width Nf = nm=32, Nf=64
    Architecture choices in Section V-A3 that determine the fidelity of FNO predictions fed to STLSQ; not fit to data but affect the downstream symbolic recovery.
  • Candidate library composition = Terms such as |v_dot|v, |v_dot|d, v_dot, |d|, and their products
    The set of basis functions in Theta is a modeling choice that essentially encodes the known hysteresis structures; the central claim depends on the true terms being present.
assumptions (3)
  • domain assumption The hysteresis operator P can be approximated by a sparse first-order ODE in displacement (and possibly a latent variable y) using the pre-defined library Theta.
    Introduced in Section II-C and IV-B; this is the modeling premise of sparse model discovery and is guaranteed in the synthetic experiments because the data are generated from such ODEs.
  • domain assumption The same discovered ODE remains valid for voltage fields outside the training distribution (RBF/Matern kernels).
    Used implicitly when NSO simulates the recovered ODE on unseen kernels in Section V; this holds only if the data-generating process is identical across kernels, which is true by construction.
  • ad hoc to paper In Experiments 3 and 4, the latent variable y can be identified from the observed data and included in the regression, but the paper does not describe how y is obtained.
    Section V-D and V-E report discovered ODEs containing y, yet Section IV-B defines the library over D, V, f(V) only; the existence and recovery of y is assumed without a described mechanism.
invented entities (1)
  • Latent variable y in the butterfly hysteresis models
    purpose: Models an internal, history-dependent state that together with displacement d produces butterfly-shaped voltage-displacement loops
    The paper introduces y via the synthetic ODEs of Experiments 3 and 4, but provides no falsifiable prediction or real measurement that would confirm y as a physical quantity; the recovery procedure is not described.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neuro-Symbolic Operator for Interpretable and Generalizable Characterization of Complex Piezoelectric Systems." pith.science (2026). https://pith.science/paper/GR4HSJVL

@misc{pith2026250524578,
  author       = {Pith},
  title        = {Pith review of: Neuro-Symbolic Operator for Interpretable and Generalizable Characterization of Complex Piezoelectric Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GR4HSJVL}},
  note         = {Machine review of arXiv:2505.24578}
}
read the original abstract

Complex piezoelectric systems are foundational in industrial applications. Their performance, however, is challenged by the nonlinear voltage-displacement hysteretic relationships. Efficient characterization methods are, therefore, essential for reliable design, monitoring, and maintenance. Recently proposed neural operator methods serve as surrogates for system characterization but face two pressing issues: interpretability and generalizability. State-of-the-art (SOTA) neural operators are black-boxes, providing little insight into the learned operator. Additionally, generalizing them to novel voltages and predicting displacement profiles beyond the training domain is challenging, limiting their practical use. To address these limitations, this paper proposes a neuro-symbolic operator (NSO) framework that derives the analytical operators governing hysteretic relationships. NSO first learns a Fourier neural operator mapping voltage fields to displacement profiles, followed by a library-based sparse model discovery method, generating white-box parsimonious models governing the underlying hysteresis. These models enable accurate and interpretable prediction of displacement profiles across varying and out-of-distribution voltage fields, facilitating generalizability. The potential of NSO is demonstrated by accurately predicting voltage-displacement hysteresis, including butterfly-shaped relationships. Moreover, NSO predicts displacement profiles even for noisy and low-fidelity voltage data, emphasizing its robustness. The results highlight the advantages of NSO compared to SOTA neural operators and model discovery methods on several evaluation metrics. Consequently, NSO contributes to characterizing complex piezoelectric systems while improving the interpretability and generalizability of neural operators, essential for design, monitoring, maintenance, and other real-world scenarios.

Figures

Figures reproduced from arXiv: 2505.24578 by the authors.

Figure 1
Figure 1. Results for Experiment 1: Top row: Blue curves represent the testing voltage fields sampled from Sine, RBF, and Matern52 kernels for FNO (first three columns, from left to right) and NSO (last three columns, from left to right). Bottom row: Hysteresis responses for FNO (first three columns, from left to right) and NSO (last three columns, from left to right) corresponding to the voltage fields on the Top row. Red da… view at source ↗
Figure 2
Figure 2. Results for Experiment 2: Top row: Testing voltage fields sampled from Sine, RBF, and Matern52 kernels for FNO (first three columns) and NSO (last three columns). Bottom row: Hysteresis responses for FNO (first three columns) and NSO (last three columns) corresponding to the voltage fields on the Top row. B. Experiment 1 For this experiment, three voltage input fields are generated: Sine, RBF, and Matern52. The Sine… view at source ↗
Figure 3
Figure 3. Results for Experiment 3: Top row: Testing voltage fields sampled from Sine, RBF, and Matern32 kernels for FNO (first three columns) and NSO (last three columns). Bottom row: Hysteresis responses for FNO (first three columns) and NSO (last three columns) corresponding to the voltage fields on the Top row [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Results for Experiment 4: Top row: Testing voltage fields sampled from Sine, RBF, and Matern32 kernels for FNO (first three columns) and NSO (last three columns). Bottom row: Hysteresis responses for FNO (first three columns) and NSO (last three columns) corresponding …
Figure 5
Figure 5. Figure 5: Top row: Displacement profiles with 20% Gaussian noise on Sine, RBF and Matern52 kernels; Bottom row: NSO predictions under corresponding noisy data [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Top row: Low-fidelity displacement profiles with 20 points each for Sine, RBF, and Matern52 kernels; Bottom row: NSO predictions under corresponding data. Among the three approaches, NSO achieves the most ac￾curate model structure, effectively capturing the dominant ph…
Figure 7
Figure 7. Figure 7: NSO predictions with Top: λ = 0.1; Bottom: λ = 0.001, on Sine, RBF, and Matern32 kernels. TABLE III: Performance of NSO for varying λ. Kernel Metric λ=0.1 λ=0.01 λ=0.001 Sine R 4.62e-2 3.91e-2 2.72e-2 RMSE 2.61e-2 2.21e-2 1.54e-2 MAE 1.92e-2 1.63e-2 1.07e-2 RBF R 4.18e…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 31 canonical work pages

  1. [1]

    Piezoelec- tric materials for controlling electro-chemical processes,

    W. Qian, W. Yang, Y . Zhang, C. R. Bowen, and Y . Yang, “Piezoelec- tric materials for controlling electro-chemical processes,” Nano-Micro Letters, vol. 12, pp. 1–39, 2020

    work page 2020

  2. [2]

    Piezoelectric materials for sus- tainable building structures: Fundamentals and applications,

    J. Chen, Q. Qiu, Y . Han, and D. Lau, “Piezoelectric materials for sus- tainable building structures: Fundamentals and applications,” Renewable and Sustainable Energy Reviews , vol. 101, pp. 14–25, 2019

    work page 2019

  3. [3]

    A review on applications of piezoelectric materials in aerospace industry,

    H. Elahi, K. Munir, M. Eugeni, M. Abrar, A. Khan, A. Arshad, and P. Gaudenzi, “A review on applications of piezoelectric materials in aerospace industry,” Integrated Ferroelectrics, vol. 211, no. 1, pp. 25– 44, 2020

    work page 2020

  4. [4]

    Applications of piezoelectric materials in biomedical engineering,

    Z. Wang, Y . Wang, C. Yang, T. Zheng, R. Luo, and Y . Wang, “Applications of piezoelectric materials in biomedical engineering,” Macromolecular Bioscience, p. 2500033, 2025

    work page 2025

  5. [5]

    A review on piezoelectric materials and their appli- cations,

    M. C. Sekhar, E. Veena, N. S. Kumar, K. C. B. Naidu, A. Mallikarjuna, and D. B. Basha, “A review on piezoelectric materials and their appli- cations,” Crystal Research and Technology , vol. 58, no. 2, p. 2200130, 2023

    work page 2023

  6. [6]

    Hysteresis in piezoelectric and ferroelectric materials,

    D. Damjanovic, “Hysteresis in piezoelectric and ferroelectric materials,” Science of hysteresis , vol. 3, pp. 337–465, 2006

    work page 2006

  7. [7]

    Adaptive estimated inverse output-feedback quantized control for piezoelectric positioning stage,

    X. Zhang, Y . Wang, C. Wang, C.-Y . Su, Z. Li, and X. Chen, “Adaptive estimated inverse output-feedback quantized control for piezoelectric positioning stage,” IEEE Transactions on Cybernetics , vol. 49, no. 6, pp. 2106–2118, 2018

    work page 2018

  8. [8]

    Piezoelectric hysteresis analysis and loss separation,

    G. Robert, D. Damjanovic, and N. Setter, “Piezoelectric hysteresis analysis and loss separation,” Journal of Applied Physics, vol. 90, no. 9, pp. 4668–4675, 2001

    work page 2001

Show all 40 references
  1. [9]

    ¨Uber die magnetische nachwirkung,

    F. Preisach, “ ¨Uber die magnetische nachwirkung,” Zeitschrift f¨ur Physik, vol. 94, no. 5, pp. 277–302, 1935

    1935

  2. [10]

    N. S. Ottosen and M. Ristinmaa, The mechanics of constitutive modeling. Elsevier, 2005

    2005

  3. [11]

    Mag- netic hysteresis modeling with neural operators,

    A. Chandra, B. Daniels, M. Curti, K. Tiels, and E. A. Lomonova, “Mag- netic hysteresis modeling with neural operators,” IEEE Transactions on Magnetics, 2024

    2024

  4. [12]

    A survey on hysteresis modeling, identification and control,

    V . Hassani, T. Tjahjowidodo, and T. N. Do, “A survey on hysteresis modeling, identification and control,” Mechanical Systems and Signal Processing, vol. 49, no. 1-2, pp. 209–233, 2014

    2014

  5. [13]

    Deep learning,

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, 2015

    2015

  6. [14]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  7. [15]

    Learning phrase representations using RNN encoder-decoder for statistical machine translation,

    K. Cho, B. Van Merri ¨enboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, and Y . Bengio, “Learning phrase representations using RNN encoder-decoder for statistical machine translation,” arXiv preprint arXiv:1406.1078, 2014

    2014 arXiv

  8. [16]

    Adaptive neural control for hysteretic nonlinear systems with hysteresis neural direct inverse compensator and its application,

    X. Zhang, Z. Hu, Y . Wang, F. Guo, Z. Li, and C.-Y . Su, “Adaptive neural control for hysteretic nonlinear systems with hysteresis neural direct inverse compensator and its application,” IEEE Transactions on Cybernetics, 2024

    2024

  9. [17]

    Generalizable models of magnetic hys- teresis via physics-aware recurrent neural networks,

    A. Chandra, T. Kapoor, B. Daniels, M. Curti, K. Tiels, D. M. Tar- takovsky, and E. A. Lomonova, “Generalizable models of magnetic hys- teresis via physics-aware recurrent neural networks,” Computer Physics Communications, p. 109650, 2025

    2025

  10. [18]

    Characterizing nonlinear piezoelectric dynamics through deep neural operator learning,

    A. Chandra, T. Kapoor, M. Curti, K. Tiels, and E. A. Lomonova, “Characterizing nonlinear piezoelectric dynamics through deep neural operator learning,” Applied Physics Letters , vol. 125, no. 26, 2024

    2024

  11. [19]

    Neural operator: Learning maps be- tween function spaces with applications to PDEs,

    N. Kovachki, Z. Li, B. Liu, K. Azizzadenesheli, K. Bhattacharya, A. Stuart, and A. Anandkumar, “Neural operator: Learning maps be- tween function spaces with applications to PDEs,” Journal of Machine Learning Research, vol. 24, no. 89, pp. 1–97, 2023

    2023

  12. [20]

    Neural operators for accelerating scientific simu- lations and design,

    K. Azizzadenesheli, N. Kovachki, Z. Li, M. Liu-Schiaffini, J. Kossaifi, and A. Anandkumar, “Neural operators for accelerating scientific simu- lations and design,” Nature Reviews Physics, vol. 6, no. 5, pp. 320–328, 2024

    2024

  13. [21]

    Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators,

    L. Lu, P. Jin, G. Pang, Z. Zhang, and G. E. Karniadakis, “Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators,” Nature Machine Intelligence , vol. 3, no. 3, pp. 218–229, 2021

    2021

  14. [22]

    Fourier neural operator for parametric partial differential equations,

    Z. Li, N. B. Kovachki, K. Azizzadenesheli, K. Bhattacharya, A. Stuart, A. Anandkumar et al. , “Fourier neural operator for parametric partial differential equations,” in International Conference on Learning Repre- sentations, 2021

    2021

  15. [23]

    Convolutional neural operators for robust and accurate learning of PDEs,

    B. Raonic, R. Molinaro, T. De Ryck, T. Rohner, F. Bartolucci, R. Alai- fari, S. Mishra, and E. de B ´ezenac, “Convolutional neural operators for robust and accurate learning of PDEs,” Advances in Neural Information Processing Systems, vol. 36, pp. 77 187–77 200, 2023

    2023

  16. [24]

    Discovery of sparse hysteresis models for piezoelectric materials,

    A. Chandra, B. Daniels, M. Curti, K. Tiels, E. A. Lomonova, and D. M. Tartakovsky, “Discovery of sparse hysteresis models for piezoelectric materials,” Applied Physics Letters , vol. 122, no. 21, 2023

    2023

  17. [25]

    Identifi- cation of Bouc–Wen hysteretic systems based on a joint optimization approach,

    M. Lin, C. Cheng, G. Zhang, B. Zhao, Z. Peng, and G. Meng, “Identifi- cation of Bouc–Wen hysteretic systems based on a joint optimization approach,” Mechanical Systems and Signal Processing , vol. 180, p. 109404, 2022

    2022

  18. [26]

    Learning the solution operator of parametric partial differential equations with physics-informed deep- onets,

    S. Wang, H. Wang, and P. Perdikaris, “Learning the solution operator of parametric partial differential equations with physics-informed deep- onets,” Science Advances, vol. 7, no. 40, p. eabi8605, 2021

    2021

  19. [27]

    Beyond accu- racy: Ecol2 metric for sustainable neural PDE solvers,

    T. Kapoor, A. Chandra, A. Stamou, and S. J. Roberts, “Beyond accu- racy: Ecol2 metric for sustainable neural PDE solvers,” arXiv preprint arXiv:2505.12556, 2025

    2025 arXiv

  20. [28]

    Neural operator: Is data all you need to model the world? an insight into the impact of physics informed machine learning,

    H. Viswanath, M. A. Rahman, A. Vyas, A. Shor, B. Medeiros, S. Her- nandez, S. E. Prameela, and A. Bera, “Neural operator: Is data all you need to model the world? an insight into the impact of physics informed machine learning,” arXiv preprint arXiv:2301.13331 , 2023

    2023 arXiv

  21. [29]

    Discovering governing equations from data by sparse identification of nonlinear dynamical systems,

    S. L. Brunton, J. L. Proctor, and J. N. Kutz, “Discovering governing equations from data by sparse identification of nonlinear dynamical systems,” Proceedings of the National Academy of Sciences , vol. 113, no. 15, pp. 3932–3937, 2016

    2016

  22. [30]

    Discovering symbolic models from deep learn- ing with inductive biases,

    M. Cranmer, A. Sanchez Gonzalez, P. Battaglia, R. Xu, K. Cranmer, D. Spergel, and S. Ho, “Discovering symbolic models from deep learn- ing with inductive biases,” Advances in Neural Information Processing Systems, vol. 33, pp. 17 429–17 442, 2020

    2020

  23. [31]

    Why are some hysteresis loops shaped like a butterfly?

    B. Drin ˇci´c, X. Tan, and D. S. Bernstein, “Why are some hysteresis loops shaped like a butterfly?” Automatica, vol. 47, no. 12, pp. 2658–2664, 2011

    2011

  24. [32]

    Fourier neural operators for arbitrary resolution climate data downscaling,

    Q. Yang, A. Hernandez-Garcia, P. Harder, V . Ramesh, P. Sattigeri, D. Szwarcman, C. D. Watson, and D. Rolnick, “Fourier neural operators for arbitrary resolution climate data downscaling,” Journal of Machine Learning Research, vol. 25, no. 420, pp. 1–30, 2024

    2024

  25. [33]

    Sound propagation in realistic interactive 3D scenes with parameterized sources using deep neural operators,

    N. Borrel-Jensen, S. Goswami, A. P. Engsig-Karup, G. E. Karniadakis, and C.-H. Jeong, “Sound propagation in realistic interactive 3D scenes with parameterized sources using deep neural operators,” Proceedings of the National Academy of Sciences , vol. 121, no. 2, p. e2312159120, 2024

    2024

  26. [34]

    Data-driven discovery of partial differential equations,

    S. H. Rudy, S. L. Brunton, J. L. Proctor, and J. N. Kutz, “Data-driven discovery of partial differential equations,” Science Advances , vol. 3, no. 4, p. e1602614, 2017

    2017

  27. [35]

    Regression shrinkage and selection via the lasso,

    R. Tibshirani, “Regression shrinkage and selection via the lasso,” Jour- nal of the Royal Statistical Society Series B: Statistical Methodology , vol. 58, no. 1, pp. 267–288, 1996

    1996

  28. [36]

    Ensemble- SINDy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and control,

    U. Fasel, J. N. Kutz, B. W. Brunton, and S. L. Brunton, “Ensemble- SINDy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and control,” Proceedings of the Royal Society A , vol. 478, no. 2260, p. 20210904, 2022

    2022

  29. [37]

    Derivative-based SINDy (DSINDy): Address- ing the challenge of discovering governing equations from noisy data,

    J. Wentz and A. Doostan, “Derivative-based SINDy (DSINDy): Address- ing the challenge of discovering governing equations from noisy data,” Computer Methods in Applied Mechanics and Engineering , vol. 413, p. 116096, 2023

    2023

  30. [38]

    Dynamic hysteresis model of grain-oriented ferromagnetic material using neural operators,

    Z. Guo, B. H. Nguyen, H. Hamzehbahmani, and R. V . Sabariego, “Dynamic hysteresis model of grain-oriented ferromagnetic material using neural operators,” arXiv preprint arXiv:2504.04863 , 2025

    2025

  31. [39]

    Adam: A method for stochastic optimization,

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

    2014 arXiv

  32. [40]

    C. K. Williams and C. E. Rasmussen, Gaussian processes for machine learning. MIT press Cambridge, MA, 2006, vol. 2, no. 3

    2006

Pith tools

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