Pith. sign in

REVIEW 4 major objections 5 minor 44 references

Polarization-Conditioned Fourier-enhanced DeepONet for Electric Field Reconstruction from EFISH Measurements

T0 review · 4 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 PC-FDON reconstructs both vertical and horizontal electric field profiles from EFISH signals with one polarization-conditioned operator network, and its dropout-based confidence metric flags measurements outside the trained profile…

desk verdict Useful incremental extension of the authors' DDON to both polarizations and variable phase mismatch; the OOD-flag validation is the main soft spot, but the paper is honest and worth reviewing. read the letter →

arxiv 2608.05937 v1 pith:KOQVIW6Q submitted 2026-08-06 physics.plasm-ph

Zhijian Yang , Edwin Setiadi Sugeng , Yaqi Zhang , Anbang Sun , Tat Loon Chng This is my paper · ORCID
classification physics.plasm-ph
keywords ElectricfieldinducedsecondharmonicgenerationEFISHPolarization-conditionedDeepoperatornetworkOperator-learningMachinelearningPhysics-informedreconstruction
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

Electric-field-induced second-harmonic generation (EFISH) measures a plasma electric field only through a coherent line-of-sight integral, so recovering the field profile is an ill-posed inverse problem. This paper claims that a single operator-learning network, the Polarization-Conditioned Fourier-enhanced DeepONet (PC-FDON), can solve that inverse problem for both vertical and horizontal field components at once, across the phase-mismatch range $u\in[-0.01,-1]$, by conditioning the reconstruction on the signal polarization and on optical parameters. It also claims that Monte-Carlo dropout gives pointwise epistemic uncertainty whose profile-level exceedance fraction reliably flags inputs outside the training distribution, such as surface dielectric barrier discharge profiles. If these claims hold, one trained model plus an uncertainty check replaces separate polarization-specific solvers and tells an experimentalist when a measurement cannot be trusted.

What carries the argument

The central object is the dimensionless EFISH forward integral $$$P^{{(2\omega)}}$_{\mathrm{norm}}(z'_o) \propto \left|\int_{-\infty}^{\infty} \frac{E'_{\mathrm{ext}}(z'-z'_o)\,$e^{{iuz'}}$}{1+iz'}\,dz'\right|^2,$$ which maps the field profile to the measured signal and whose kernel carries the Gouy phase $1/(1+iz')$ and the wave-vector mismatch $e^{iuz'}$. PC-FDON learns the inverse of this integral as an operator: a Fourier-enhanced branch extracts global spectral features, a FiLM-conditioned polarization branch modulates those features according to the polarization label, a trunk network evaluates the operator at query positions, and a physics-informed loss enforces that the predicted field, when pushed back through the forward integral, reproduces the measured signal. The MC-dropout procedure turns the same trained network into a Bayesian approximation whose pointwise standard deviation, compared against validation percentile thresholds via the exceedance fraction of Eq. (11), yields the paper's confidence levels.

What would settle it

Generate a synthetic EFISH signal from a clearly asymmetric, single-peaked vertical field profile inside the trained phase-mismatch range $u\in[-0.01,-1]$, feed it to PC-FDON, and compare the reconstruction against the known ground truth: if the error stays below the paper's $\epsilon_{\mathrm{MSE}}=4\times10^{-3}$ threshold and the exceedance fraction remains in-distribution, the stated shape limitation is false; if the error is large or the OOD flag fires, the limitation holds as stated.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that EFISH inversion can be performed by one physics-informed operator network rather than by polarization-specific networks tuned to a single phase mismatch. The network takes a normalized EFISH signal profile, the phase-mismatch parameter $u$, and a binary polarization label $\psi$, and outputs the normalized electric field profile $E'_{\mathrm{ext}}(z')$ together with a predicted polarization class. Fourier layers give the branch a spectral inductive bias for the Gouy-phase and wave-vector-mismatch part of the forward kernel; FiLM and a gated unit modulate the shared features by polarization; and a physics-informed loss re-inserts the predicted field into the forward integral of Eq. (3) to enforce self-consistency. Trained on symmetric bell-shaped and double-peaked vertical profiles and bipolar horizontal profiles at seven discrete $u$ values, the model reconstructs noise-free, incomplete, and noisy inputs, generalizes to unseen Quartic and Raised Cosine families at accuracy comparable to the earlier polarization-specific model, and its MC-dropout exceedance fraction flags SDBD profiles as out-of-distribution while accepting experimental electrostatic measurements.

Load-bearing premise

The load-bearing premise is that the true electric field profile belongs to the trained shape families: symmetric bell-shaped or double-peaked vertical profiles and bipolar horizontal profiles; the paper states in Section 3.3.3 that asymmetric, triple-peaked, or otherwise non-smooth profiles remain outside the model's scope and are at best caught by the out-of-distribution flag rather than reconstructed correctly.

Editorial extensions

If this is right

  • One trained PC-FDON replaces separate vertical- and horizontal-polarization reconstruction models; running it once with each polarization label yields both components of the electric field vector.
  • Across the phase-mismatch range $u\in[-0.01,-1]$, the model reconstructs profiles without retuning for each optical setup, covering the common 1064 nm and 532 nm configurations listed in the paper.
  • The location-dependent exceedance fraction gives a profile-level confidence label (very high, high, borderline, low) that flags measurements outside training, so a user can reject a reconstruction instead of trusting it blindly.
  • The physics-informed loss plus forward-signal comparison allows validation against the measured EFISH profile when no ground-truth field is available, which is the normal experimental situation.
  • On unseen Quartic and Raised Cosine profile families at SNR=20 dB, the unified model stays comparable to the previous polarization-specific model, with the paper attributing the small gap to the halved vertical-polarization training budget.

Reading between the lines

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

  • As an editorial extension, the OOD flag certifies similarity to the training manifold, not physical correctness, so a user who sees a low-confidence label still needs the paper's recommended noise-level check to decide whether the field is genuinely new or the measurement is just noisy.
  • A testable extension is to add asymmetric vertical and multi-peaked profile families to the training set; if the Fourier layers and physics-informed loss then reconstruct them accurately, the symmetry assumption in Section 3.3.3 would be shown to be a data limitation rather than an architectural one.
  • The same polarization-conditioning and Fourier-bias recipe should transfer to other line-of-sight-integrated plasma diagnostics, such as emission spectroscopy, because those diagnostics share the structure of an integral kernel modulated by a discrete experimental parameter.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces PC-FDON, a DeepONet-style operator network for reconstructing electric field profiles from EFISH measurements. The architecture combines a Fourier-enhanced branch, a FiLM/gated polarization-conditioning branch, an auxiliary polarization classifier, and a physics-informed loss term. Training data are synthetic EFISH profiles derived from the forward integral in Eq. (3), covering vertical (bell-shaped/double-peak) and horizontal (bipolar) field families over seven discrete phase-mismatch values. The authors report reconstruction rates of 99.6%, 94.3%, and 95.1% below an MSE threshold of 0.004 for noise-free, incomplete, and noisy inputs, respectively; generalization tests on Quartic and Raised Cosine families show performance comparable to their earlier DDON. An MC-dropout uncertainty estimate is summarized by location-dependent exceedance fractions, and this OOD flag is applied to two simulated SDBD configurations, which are flagged as out-of-distribution, and to experimental disk-disk and sphere-sphere fields, which are classified as in-distribution. The paper explicitly acknowledges in Section 3.3.3 that the model remains limited to symmetric bell-shaped/double-peak vertical profiles and bipolar horizontal profiles.

Significance. If the results hold, PC-FDON would be a practically useful unified inverse solver for EFISH: one trained operator network would handle both polarization components and a range of optical parameters, with a built-in confidence/OOD indicator. The paper is strong in engineering detail: the architecture, hyperparameters, dataset sizes, and training schedule are specified in Tables A.1 and B.1, and the synthetic performance numbers are clearly quantified. The decision to state the profile-shape limitation explicitly is also a strength. However, the claimed advances are not yet fully evidenced: the OOD flag rests on a very small validation set without false-positive/false-negative controls, and the three architectural innovations are never isolated by ablations. The synthetic validation and the physics-informed loss also share the same forward model, so the reported MSEs primarily demonstrate self-consistency with Eq. (3) rather than absolute accuracy against independent field measurements.

major comments (4)
  1. [§3.3.1 and Eq. (12)] The OOD detection claim is validated only on two SDBD cases, with no hard in-distribution controls and no measured false-positive or false-negative rates. Since Section 3.3.3 concedes that reconstruction is limited to specific profile shapes, the OOD flag is the only safeguard against silent failure on real discharge profiles. The authors should report performance on a controlled test set that includes difficult but in-distribution inputs (e.g., extreme u values, high noise, heavy cropping, high-curvature in-family profiles) and genuinely out-of-family shapes, and should quantify the trade-off between the p95/p99 thresholds in Eq. (12).
  2. [§2.5 and §3.3.2] The exceedance-fraction metric is calibrated on validation-set percentiles but its relationship to actual reconstruction error is never demonstrated. The paper shows that OOD flags can arise from two distinct causes, noise-induced and shape-induced, but provides no quantitative separation of these causes. A calibration analysis, such as a reliability diagram or a plot of exceedance fraction versus true MSE on held-out in-distribution and OOD sets, is needed to support the claim that the metric reflects model confidence rather than merely input difficulty.
  3. [§2.3 and §3.2] The paper claims three targeted advances, Fourier enhancement, polarization conditioning via FiLM/gating, and a physics-informed loss, but no ablation isolates their contributions. The comparison in Section 3.2 is only PC-FDON versus the authors' previous DDON, so it cannot attribute any observed difference to the Fourier branch, the FiLM/gated mechanism, or the PINN term. I recommend adding ablation variants, each removing one component while keeping the training data and evaluation protocol identical, and reporting the resulting MSE distributions and OOD behavior.
  4. [Eq. (3) and §2.1] The training data, the physics-informed loss, and the synthetic test sets all use the same forward model, Eq. (3). Consequently, the reported fractions below MSE = 0.004 establish that the network has learned the inverse of the assumed forward operator, but they do not by themselves validate the model against the real EFISH physics. The paper should state this limitation explicitly and discuss how model-form error in Eq. (3) could affect field reconstruction, in addition to the already-listed experimental limitations.
minor comments (5)
  1. [Eq. (9)] The loss terms are written without sample indices; for reproducibility, define the per-sample MSE and BCE operations explicitly over the training batch.
  2. [Figure 2 caption] The caption reports SNR = 20 dB for the noisy condition, while the jitter-layer augmentation in Section 2.1 uses SNR ≈ 26 dB; clarify whether the test noise level is intentionally different from the training augmentation level.
  3. [Section 3.3.2] The phrase 'the model’s confidence is well-placed' is not quantitatively supported; a comparison of uncertainty estimates with the actual reconstruction error would make this statement precise.
  4. [Table A.1] The 'Num' column for layer counts is difficult to parse; including a small legend or explicitly listing the number of repetitions for each block would improve reproducibility.
  5. [Section 2.1.2] Equation (5) mixes dimensional quantities (z, y, l) with a dimensionless acosh term; since all profiles are later peak-normalized, state clearly which variables are in units of z_R before normalization.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the reconstruction is tested on held-out and unseen families; the physics-informed loss is a consistency regularizer, not a fitted input.

full rationale

The core inverse mapping is trained on synthetic pairs generated by inserting diverse profile families into the standard forward EFISH integral (Eq. 3), and evaluated on a held-out split, two unseen function families (Quartic, Raised Cosine), SDBD simulations, and experimental electrostatic cases. None of these targets is a refit of a training quantity, so the central reconstruction claim does not reduce to its inputs by construction. The physics-informed term (Eq. 9) penalizes mismatch between the input EFISH profile and the forward projection of the predicted field; this is a regularizer enforcing self-consistency with the same governing equation used to generate data, not an independent physical law, but it is not a circular prediction because the network is not given the field as an input. The polarization-classification head predicts the label psi that is also supplied as an input; that auxiliary task is redundant as a diagnostic but does not carry the reconstruction claim. The OOD thresholds (Eq. 12) are calibrated on the validation set and demonstrated on two SDBD cases without hard in-distribution controls, a validation gap rather than circularity. Several citations are to the authors' prior work [9,14], but they are used for dataset generation and baseline comparison, not as load-bearing uniqueness or ansatz authority. Score 2 reflects minor self-consistency and self-citation, not a derived-equivalence circularity.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the correctness of the EFISH forward model, the representativeness of the synthetic training families, and the validity of MC dropout as an uncertainty approximation. These are standard domain assumptions rather than demonstrated results. The free parameters are training hyperparameters chosen by hand or validation tuning; they do not enter a physical derivation.

free parameters (7)
  • physics loss weight lambda_PINN = 0.05
    Chosen by validation during parametric optimization (Table A.1); controls strength of the forward-model consistency term in Eq. (9).
  • classification loss weight lambda_CLS = 0.5
    Balances the auxiliary polarization classification head against reconstruction losses (Table A.1).
  • dropout rate = 0.1 default, 0.05 predictor layer
    Applied at inference for MC dropout; directly shapes the uncertainty estimates in Eq. (10).
  • Fourier modes per layer = 256, 128, 32, 16
    Number of spectral modes in the Fourier sub-branch, selected after parametric optimization (Table A.1).
  • initial learning rate = 1e-3
    Adam optimizer setting with reduce-on-plateau scheduler, chosen for stable training (Section 2.4).
  • MC dropout passes N = 100
    Number of stochastic forward passes used for uncertainty; authors state this provides stable standard deviation estimates (Section 2.5).
  • batch size = 512
    Training batch size chosen during hyperparameter tuning (Table A.1).
assumptions (6)
  • domain assumption The forward EFISH model, Eq. (3), accurately describes measured signals for both polarizations, including uniform alpha^(3), N, and Delta k along the beam.
    Invoked throughout data generation and the physics-informed loss; the paper itself lists non-uniform gas properties as outside scope in Section 3.3.3.
  • domain assumption The synthetic training families (symmetric Fuzzy and Voigt vertical profiles, bipolar horizontal profiles) are representative of real field distributions to which the model will be applied.
    Data generation in Section 2.1 relies on these families; Section 3.3.3 acknowledges the model is limited to these profile shapes.
  • domain assumption MC dropout with frozen dropout approximates the Bayesian posterior predictive distribution of a deep Gaussian process, so its standard deviation is a meaningful epistemic uncertainty.
    Section 2.5 adopts the Gal and Ghahramani interpretation of dropout as Bayesian approximation; no independent calibration of the uncertainty is provided.
  • domain assumption Peak-normalizing EFISH signals and field profiles preserves the shape information needed for inversion, with absolute amplitude recovered separately via calibration.
    Section 2.1 states all profiles and signals are max-normalized and that calibration is used to recover E_o.
  • domain assumption COMSOL electrostatic solutions used as ground truth for the SDBD and experimental benchmarks approximate the true electric fields in those configurations.
    Appendix B describes solving Eq. (B.1) with boundary conditions; the experimental comparisons in Section 3.3.2 use the same simulated fields as reference.
  • domain assumption DeepONet universal approximation and Fourier spectral convolutions provide sufficient capacity and inductive bias to represent the inverse EFISH operator over the chosen function families.
    Architecture choices in Section 2.3 assume the operator learning framework can capture the inverse mapping; proven only empirically on the selected test sets.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Polarization-Conditioned Fourier-enhanced DeepONet for Electric Field Reconstruction from EFISH Measurements." pith.science (2026). https://pith.science/paper/KOQVIW6Q

@misc{pith2026260805937,
  author       = {Pith},
  title        = {Pith review of: Polarization-Conditioned Fourier-enhanced DeepONet for Electric Field Reconstruction from EFISH Measurements},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KOQVIW6Q}},
  note         = {Machine review of arXiv:2608.05937}
}
read the original abstract

Electric-field-induced second-harmonic generation (EFISH) is an established laser diagnostic for quantifying electric fields in plasmas, yet ensuring field accuracy remains challenging given the coherent, path-integrated nature of the signal. We address this via machine learning, developing a Polarization-Conditioned Fourier-enhanced Deep Operator Network (PC-FDON) -- a unified operator-learning model that reconstructs field profiles from EFISH measurements across both polarizations and various optical parameters. Its architecture incorporates three advances: (i) a Fourier-enhanced branch providing inductive bias for the Gouy phase shift and wave-vector mismatch; (ii) a polarization-conditioning branch encoding signal polarization via Feature-wise Linear Modulation (FiLM) and gated units, enabling one model to handle both polarizations; and (iii) a physics-informed loss enforcing self-consistency with the governing EFISH equation. Trained on data spanning multiple function families, polarization states, and phase-mismatch values, PC-FDON achieves promising reconstruction under noise-free, incomplete, and noisy inputs, with generalizability comparable to our previous polarization-specific model. Pointwise epistemic uncertainty estimates via Monte Carlo dropout reflect model confidence and enable out-of-distribution (OOD) detection through a location-dependent exceedance fraction metric. Validation is performed on realistic electrode configurations under both polarizations and varying Rayleigh ranges, including a simulated surface dielectric barrier discharge where the framework correctly flags OOD inputs, and experimental data showing good agreement with simulations. The architecture -- spectral inductive bias, conditional modulation, and dataset-specific uncertainty -- shows strong potential for broader application beyond plasma diagnostics.

Figures

Figures reproduced from arXiv: 2608.05937 by the authors.

Figure 1
Figure 1. PC-FDON architecture for operator learning: inputs ({ 𝑃 2𝜔 norm (𝑧 ′ o ), 𝑢′ } for EFISH branch, {𝜓} for polarization branch, and { 𝑧 ′ o } for trunk net) and outputs (electric field, 𝐸′ ext(𝑧 ′ ) and polarization class). Further details are listed in the Table A.1 in Appendix A. mismatch value 𝑢 = −0.068, and is unable to handle the horizontal polarization cases with a bipolar character as well as vertically polari… view at source ↗
Figure 2
Figure 2. Polarization-conditioned reconstruction performance of PC-FDON under three input conditions: (a–c) noise-free, (d–f) incomplete, and (g–j) noisy (SNR = 20 dB). Predictions are shown for both the vertical (𝐸′ 𝑦 , panels i–ii) and horizontal (𝐸′ 𝑥 , panels iii–iv) polarization states. For each condition, the first column (a, d, g) displays the polarization classification accuracy via a confusion matrix, the second col… view at source ↗
Figure 3
Figure 3. Generalization performance of PC-FDON (orange) and DDON [14] (purple) on two unseen function families under noisy conditions (SNR = 20 dB): (a,b,e,f) Quartic and (c,d,g,h) Raised Cosine. Panels (a–d) show PC-FDON predictions; panels (e–h) show DDON predictions. Both models are evaluated only for the vertically polarized field component, as the earlier DDON model was designed for predicting only that specific (vertic… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Predicted electric field components (horizontal: 𝐸𝑥,P ; vertical: 𝐸𝑦,P ) and associated uncertainty quantification for (a,b,e,f) disk and (c,d,g,h) gear SDBD configurations. Panels (a,e) and (c,g) show the reconstructed field profiles benchmark against the simulation g…
Figure 5
Figure 5. Figure 5: Experimental electrostatic field reconstruction via PC-FDON for (a, c) disk–disk and (b, d) sphere–sphere electrode configurations. Panels (a, b) show the reconstructed vertical field profile 𝐸′ 𝑦 and panels (c, d) show the horizontal field profile 𝐸′ 𝑥 , both benchmar…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 40 canonical work pages

  1. [1]

    Dogariu, B

    A. Dogariu, B. M. Goldberg, S. O’Byrne, R. B. Miles, Species-Independent Femtosecond Localized Electric Field Measurement, Phys. Rev. Appl. 7 (2017) 024024

    work page 2017

  2. [2]

    I. J. Bigio, R. S. Finn, J. F. Ward, Electric-Field Induced Harmonic Generation as a Probe of the Focal Region of a Laser Beam, Appl. Opt., AO 14 (1975) 336–342

    work page 1975

  3. [3]

    T. L. Chng, S. M. Starikovskaia, M.-C. Schanne-Klein, Electric field measurements in plasmas: How focusing strongly distorts the E-FISH signal, Plasma Sources Sci. Technol. 29 (2020) 125002

    work page 2020

  4. [4]

    T.L.Chng,D.Z.Pai,O.Guaitella,S.M.Starikovskaia,A.Bourdon,EffectoftheelectricfieldprofileontheaccuracyofE-FISHmeasurements in ionization waves, Plasma Sources Sci. Technol. 31 (2022) 015010

    work page 2022

  5. [5]

    Raskar, K

    S. Raskar, K. Orr, I. V. Adamovich, T. L. Chng, S. M. Starikovskaia, Spatially enhanced electric field induced second harmonic (SEEFISH) generation for measurements of electric field distributions in high-pressure plasmas, Plasma Sources Sci. Technol. 31 (2022) 085002

    work page 2022

  6. [6]

    Hogue, P

    J. Hogue, P. Cusson, M. Meunier, D. V. Seletskiy, S. Reuter, Sensitive detection of electric field-induced second harmonic signals, Opt. Lett. 48 (2023) 4601–4604

    work page 2023

  7. [7]

    Vorenkamp, S

    M. Vorenkamp, S. A. Steinmetz, T. Y. Chen, Y. Ju, C. J. Kliewer, Suppression of coherent interference to electric-field-induced second- harmonic (E-FISH) signals for the measurement of electric field in mesoscale confined geometries, Opt. Lett., OL 48 (2023) 1930–1933

    work page 2023

  8. [8]

    N. D. Lepikhin, D. Luggenhölscher, U. Czarnetzki, Increasing the EFISH sensitivity by an optical parametric amplifier, Opt. Lett. 50 (2025) 5809–5812

    work page 2025

Show all 44 references
  1. [9]

    Z. Yang, E. S. Sugeng, T. L. Chng, A deep learning approach for electric field profile reconstruction based on the E-FISH method, Plasma Sources Sci. Technol. 34 (2025) 02LT01

    2025

  2. [10]

    Y. Guo, A. Limburg, J. Laarman, J. Teunissen, S. Nijdam, Measurement of the electric field distribution in streamer discharges, Phys. Rev. Res. 7 (2025) 013051

    2025

  3. [11]

    S.Chen,H.He,Y.Chen,Z.Liu,S.Xie,J.Che,K.He,W.Chen, Measurementofinhomogeneouselectricfieldbasedonelectricfield-induced second-harmonic generation, Measurement 231 (2024) 114576

    2024

  4. [12]

    Q. Luo, X. Chen, Y. Zhang, J. Xie, L. Lan, A discrete inversion measurement method for non-uniform electric fields based on electric field induced second harmonic generation, Opt. Laser Technol. 192 (2025) 113939

    2025

  5. [13]

    T.Sato,T.Umemoto,M.Sato,T.Fujii,A.Kumada,Phase-resolvedmeasurementofelectric-field-inducedsecondharmonicsanditsapplication to noninvasive electric field sensing, Plasma Sources Sci. Technol. 34 (2025) 105019

    2025

  6. [14]

    Z.Yang,E.S.Sugeng,M.Alicherif,T.L.Chng, Aninterpretableoperator-learningmodelforelectricfieldprofilereconstructionindischarges based on the EFISH method, Plasma Sources Sci. Technol. 35 (2026) 025035

    2026

  7. [15]

    M.Alicherif,E.Sugeng,Z.Yang,A.L.Deanna,T.L.Chng,Towardquantitativeelectric-fieldmeasurementsofinceptioncloudsinnanosecond discharges using E-FISH assisted by machine learning, Plasma Sources Sci. Technol. 35 (2026) 065006

    2026

  8. [16]

    X.Zheng,Z.Zhao,H.Zheng,Z.Huang,Z.Sun,J.Li,PolarizationpropertiesofE-FISHsignalsandoptimizationofsimultaneousmeasurement of electric field vectors, Plasma Sources Sci. Technol. 33 (2024) 035002

    2024

  9. [17]

    J. F. Ward, I. J. Bigio, Molecular second- and third-order polarizabilities from measurements of second-harmonic generation in gases, Phys. Rev. A 11 (1975) 60–66

    1975

  10. [18]

    R. W. Boyd, Nonlinear Optics, 4th ed ed., Elsevier Science & Technology, San Diego, 2020

    2020

  11. [19]

    T. L. Chng, M. Naphade, B. M. Goldberg, I. V. Adamovich, S. M. Starikovskaia, Electric field vector measurements via nanosecond electric- field-induced second-harmonic generation, Opt. Lett., OL 45 (2020) 1942–1945

    2020

  12. [20]

    T.L.Chng,I.S.Orel,S.M.Starikovskaia,I.V.Adamovich, Electricfieldinducedsecondharmonic(E-FISH)generationforcharacterization of fast ionization wave discharges at moderate and low pressures, Plasma Sources Sci. Technol. 28 (2019) 045004

    2019

  13. [21]

    Alicherif, E

    M. Alicherif, E. S. Sugeng, Z. Yang, D. A. Lacoste, T. L. Chng, Quantitative electric field diagnostics in plasma-flame interaction via machine-learning-assisted inverse E-FISH, Proc. Combust. Inst. 42 (2026) 106276

    2026

  14. [22]

    A. D. Bonzanini, K. Shao, D. B. Graves, S. Hamaguchi, A. Mesbah, Foundations of machine learning for low-temperature plasmas: Methods and case studies, Plasma Sources Sci. Technol. 32 (2023) 024003

    2023

  15. [23]

    T. Wang, Y. Wang, J. Zhou, B. Peng, X. Song, C. Zhang, X. Sun, Q. Niu, J. Liu, S. Chen, K. Chen, M. Li, P. Feng, Z. Bi, M. Liu, Y. Zhang, C. Fei, C. H. Yin, L. K. Yan, From Aleatoric to Epistemic: Exploring Uncertainty Quantification Techniques in Artificial Intelligence, 2025...

    2025 doi arXiv

  16. [24]

    Y. Gal, Z. Ghahramani, Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning, in: Proc. 33rd Int. Conf. Mach. Learn., PMLR, 2016, pp. 1050–1059

    2016

  17. [25]

    T. Ida, M. Ando, H. Toraya, Extended pseudo-Voigt function for approximating the Voigt profile, J Appl Cryst 33 (2000) 1311–1316

    2000

  18. [26]

    B.M.Goldberg,T.L.Chng,A.Dogariu,R.B.Miles, Electricfieldmeasurementsinanearatmosphericpressurenanosecondpulsedischarge with picosecond electric field induced second harmonic generation, Applied Physics Letters 112 (2018) 064102

    2018

  19. [27]

    S.Nakamura,M.Sogame,M.Sato,T.Fujii,A.Kumada, ElectricfieldmeasurementinDCcoronadischargeinatmosphericpressureairusing E-FISHG and laser-triggering methods, Plasma Sources Sci. Technol. 33 (2024) 055010

    2024

  20. [28]

    Taylor, G

    L. Taylor, G. Nitschke, Improving Deep Learning with Generic Data Augmentation, in: 2018 IEEE Symp. Ser. Comput. Intell. SSCI, 2018, pp. 1542–1547. doi:10.1109/SSCI.2018.8628742

    2018

  21. [29]

    Holmstrom, P

    L. Holmstrom, P. Koistinen, Using additive noise in back-propagation training, IEEE Trans. Neural Netw. 3 (1992) 24–38

    1992

  22. [30]

    An, The Effects of Adding Noise During Backpropagation Training on a Generalization Performance, Neural Comput 8 (1996) 643–674

    G. An, The Effects of Adding Noise During Backpropagation Training on a Generalization Performance, Neural Comput 8 (1996) 643–674

    1996

  23. [31]

    L. Lu, P. Jin, G. E. Karniadakis, DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators, Nat Mach Intell 3 (2021) 218–229

    2021

  24. [32]

    L.Lu,P.Jin,G.Pang,Z.Zhang,G.E.Karniadakis, LearningnonlinearoperatorsviaDeepONetbasedontheuniversalapproximationtheorem of operators, Nat Mach Intell 3 (2021) 218–229. 21

    2021

  25. [33]

    H.Park,W.Choe,S.Yoo,Spatiallyresolvedemissionusingageometry-dependentsystemfunctionanditsapplicationtoexcitationtemperature profile measurement, Spectrochimica Acta Part B: Atomic Spectroscopy 65 (2010) 1029–1032

    2010

  26. [34]

    T.Wu,T.Wu,S.Zhou,Q.Liao,P.Lu, Deeplearning-basedspatiotemporalsequenceforecastingofphysicalfieldsintindropletlaser-produced plasma, Plasma Sci. Technol. 27 (2025) 075501

    2025

  27. [35]

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, A. Anandkumar, Fourier Neural Operator for Parametric Partial Differential Equations, 2021. doi:10.48550/arXiv.2010.08895.arXiv:2010.08895

    2021 doi arXiv

  28. [36]

    S. L. Brunton, J. N. Kutz, Promising directions of machine learning for partial differential equations, Nat Comput Sci 4 (2024) 483–494

    2024

  29. [37]

    A.Cong,Y.Jin,H.Yi,Y.Jiang,J.Li,S.Wang,L.Hou, DeeponetwithamodifiedFNObranchnetworkbasedonnonuniformFouriertransform for irregular input grids, Nonlinear Dyn 114 (2026) 323

    2026

  30. [38]

    E. S. Santos, G. F. Barros, A. C. N. Oliveira, R. M. Silva, R. S. M. Freitas, D. M. Valiveti, X.-H. Wu, F. A. Rochinha, A. L. G. A. Coutinho, Hybrid DeepONet surrogates for multiphase flow in porous media, Journal of Computational Physics 561 (2026) 114981

    2026

  31. [39]

    Perez, F

    E. Perez, F. Strub, H. De Vries, V. Dumoulin, A. Courville, FiLM: Visual Reasoning with a General Conditioning Layer, AAAI 32 (2018)

    2018

  32. [40]

    Bengio, Practical recommendations for gradient-based training of deep architectures, 2012

    Y. Bengio, Practical recommendations for gradient-based training of deep architectures, 2012. doi:10.48550/arXiv.1206.5533. arXiv:1206.5533

    2012 doi arXiv

  33. [41]

    Abdar, F

    M. Abdar, F. Pourpanah, S. Hussain, D. Rezazadegan, L. Liu, M. Ghavamzadeh, P. Fieguth, X. Cao, A. Khosravi, U. R. Acharya, V. Makarenkov, S. Nahavandi, A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges, Information Fusion 76 (202...

    2021

  34. [42]

    Tousignant, P

    A. Tousignant, P. Lemaître, D. Precup, D. L. Arnold, T. Arbel, Prediction of Disease Progression in Multiple Sclerosis Patients using Deep Learning Analysis of MRI Data, in: Proc. 2nd Int. Conf. Med. Imaging Deep Learn., PMLR, 2019, pp. 483–492

    2019

  35. [43]

    M. F. Kasim, D. Watson-Parris, L. Deaconu, S. Oliver, P. Hatfield, D. H. Froula, G. Gregori, M. Jarvis, S. Khatiwala, J. Korenaga, J. Topp- Mugglestone,E.Viezzer,S.M.Vinko, Buildinghighaccuracyemulatorsforscientificsimulationswithdeepneuralarchitecturesearch, Mach. Learn.: Sci...

    2021

  36. [44]

    Zhang, Y

    Y. Zhang, Y. Guo, Y. Zhu, A. Sun, Inhibition and promotion of quasi-uniform to filamentary discharge transition in negative repetitive nanosecond surface dielectric barrier discharge, Phys. Plasmas 33 (2026) 023511. 22

    2026

Pith tools

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