Physics-Informed Deep Learning Model for Line-integral Diagnostics Across Fusion Devices

Changqing Zou; Chenshu Hu; Cong Wang; Haiping Wang; Jing Li; Renjie Yang; Weizhe Yang; Xianli Huang; Xinyao Yu; Yixiong Wei

REVIEW 4 major objections 6 minor 43 references

Physics-Informed Deep Learning Model for Line-integral Diagnostics Across Fusion Devices

T0 review · 4 major / 6 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 Physics-aware neural networks with built-in diagnostic geometry cut plasma profile reconstruction error.

desk verdict Solid wrapper architecture for line-integral tomography; the PI and Softplus results hold up reasonably, but the PILF 'corrects inversion errors' claim is an artifact of the loss definition and should be reframed. read the letter →

arxiv 2412.00087 v3 pith:ZVBNDBLQ submitted 2024-11-27 cs.LG cs.CE

Cong Wang , Weizhe Yang , Haiping Wang , Renjie Yang , Jing Li , Zhijun Wang , Yixiong Wei , Xianli Huang

show 5 more authors

Chenshu Hu Zhaoyang Liu Xinyao Yu Changqing Zou Zhifeng Zhao

This is my paper · ORCID

classification cs.LGcs.CE

keywords physics-informedneuralnetworksplasmaprofilereconstructionline-integraldiagnosticssoftX-raytomographytokamaksurrogatemodelsEASTHL-2A

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 argues that a neural network trained to invert line-integral fusion diagnostics (soft X-ray measurements on the EAST and HL-2A tokamaks) becomes substantially more accurate when the diagnostic's geometry is injected directly into the network and when a line-integral consistency term is added to the loss. The authors propose a reusable wrapper architecture, Onion, that can wrap standard backbones such as VGG and ResNet, and they report reductions in average relative reconstruction error of about $0.84\times 10^{-2}$ on synthetic datasets and about $0.06\times 10^{-2}$ on experimental datasets from the physics-information side chain alone, with further gains from a Softplus output activation and from the physics-informed loss. If these claims hold, surrogate models could reconstruct 2D plasma profiles in milliseconds rather than seconds, making real-time monitoring and between-shot analysis more practical for fusion reactors. The paper also introduces a synthetic data generator for line-integral diagnostics and releases experimental soft X-ray datasets, so the architecture can be tested and extended by others.

What carries the argument

The key machinery is the Onion architecture: a physical-information (PI) side chain that takes the diagnostic response matrices $R_i$ (path lengths of each line of sight through each pixel, arranged as $(n, z, r)$ tensors), applies positional encoding and normalization, and passes them through a simplified VGG feature extractor; the resulting features are fused with the backbone's output by element-wise multiplication, so that the network's activations are multiplicatively gated by where the diagnostic 'looks.' The second component is the physics-informed loss function (PILF), $L = w_1 \mathrm{loss}_1 + w_2 \mathrm{loss}_2 + \lambda L_2$, where $\mathrm{loss}_1$ is the MSE between reconstruction $Net(x)$ and target profile $y$, and $\mathrm{loss}_2 = \frac{1}{n}\sum_j (R_j \cdot Net(x) - x_j)^2$ enforces that the back-projection of the reconstruction matches the measured chords; the weight $w_2 = c_1 \cdot \mathrm{loss}_1/\mathrm{loss}_2$ dynamically balances the two terms. A third, smaller component is the use of Softplus (instead of ReLU) in the final two fully connected layers, which smooths edge artifacts in the reconstructed profiles. The response matrices are the load-bearing physics input: they encode the line-integration principle, Eq. (2-1): $x_i = R_i \cdot y + \Delta_i$.

What would settle it

Take a diagnostic with a known response matrix and a validation set of independently measured ground-truth profiles (e.g., from Thomson scattering rather than from an inversion algorithm); deliberately corrupt the response matrix by rotating or shifting the lines of sight by a small amount and retrain Onion: if the corrupted-physics model matches or exceeds the clean-physics model's accuracy, or if the clean-physics reconstructions disagree with the independent measurements more than a backbone without physics, then the assumed forward model is not the source of the improvement.

Watch

Extended reading notes

Core claim

The central claim is that the Onion architecture—a wrapper that fuses a positional-encoded response-matrix side chain with a backbone network through element-wise multiplication and augments the loss with a back-projection consistency term—improves the accuracy of neural-network reconstruction of 2D plasma emissivity profiles from line-integral measurements. On synthetic data with essentially zero forward-model error, adding the physics information lowered the average relative error $E_1$ by roughly $0.84\times 10^{-2}$ (about 52%), and on experimental data it lowered $E_1$ slightly while leaving the back-projection error $E_2$ almost unchanged; the physics-informed loss component then reduced $E_2$ by up to $4.59\times 10^{-2}$ on the HL-2A dataset, at the cost of larger $E_1$, because the loss pulls predictions toward agreement with the raw diagnostic inputs rather than toward the (error-containing) inversion-algorithm targets.

Load-bearing premise

The load-bearing premise is that the response matrices $R_i$ exactly describe the diagnostic's line integrals, with negligible noise $\Delta_i$; if the geometry is mis-calibrated, the physics side chain and consistency loss will pull reconstructions toward the wrong measurement model, and the paper offers no independent calibration check.

Editorial extensions

If this is right

If the architecture works as claimed, the same wrapper can be reused across fusion devices and across diagnostic types whose forward model is a known response matrix, since the backbone is interchangeable and the physics input is device-specific geometry.
Reconstruction times of a few milliseconds on GPU (and under about 60 ms on CPU) versus seconds to minutes for Fourier–Bessel or nonstationary Gaussian process tomography could enable real-time plasma control and high-throughput between-shot analysis.
The physics-informed loss can partially correct errors baked into inversion-algorithm-generated training labels, because it pulls the reconstruction's back-projection toward the raw measured signals rather than only toward the target profiles.
On noise-free synthetic data, the gains from physics injection are large ($E_1$ down by roughly 52–71%), but on experimental data the gains shrink to about 15–27%, suggesting that dataset noise (the inherent $\bar{\varepsilon}$ error) is the main limiter of further improvement.

Reading between the lines

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

If response matrices for other line-integral diagnostics (bolometry, interferometry, neutron cameras) are known, the same Onion wrapper should transfer directly, making it a candidate standard module for diagnostic surrogates.
The trade-off between $E_1$ and $E_2$ controlled by the hyperparameter $c_1$ suggests a tunable operating point: for physics-faithful reconstructions one may accept larger deviation from inversion-algorithm labels, and an independent calibration dataset (e.g., from Thomson scattering) could select the optimal $c_1$.
The lack of uncertainty information in the experimental datasets is a stated limitation; an immediate testable extension is to train Onion on synthetic data with injected noise and uncertainty weights to see whether the physics-side-chain benefits persist under realistic noise levels.
The Softplus gain on synthetic data ($E_1$ down about 71%) versus experimental data (about 27%) hints that edge smoothness is closely tied to noise-free labels; adding an edge-regularization term, which the authors mention as future work, might recover some of the experimental gap.

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.

Desk Editor's Note

read the letter

I'll give you the short version first: the Onion architecture is a genuinely useful wrapper for neural line-integral tomography, and the PI-injection and Softplus improvements are plausible and worth trying. But the paper's headline claim that the physics-informed loss (PILF) reduces inversion-algorithm errors is not supported. The E2 drop in Table 5 is forced—loss2 in Eq. (3-5) penalizes exactly what E2 in Eq. (4-2) measures—and the same rows show E1 getting substantially worse (from 0.50 to 1.35e-2 on Exp_EAST for VggOnion_PI, and from 0.21 to 1.94e-2 on Exp_HL-2A). Claiming that this corrects inversion errors requires either an independent ground truth or a validated forward model. Neither is provided. The response matrices are taken as given, with no calibration check; the experimental targets have known inconsistency with the inputs (that's the whole motivation); and the only datasets with true ground truth are noiseless, where loss1 alone already works. So the PILF section over-reaches.

What is genuinely new: the Onion wrapper, the multiplication-based PI side chain, and the systematic comparison across four datasets with two backbones. The synthetic data model is simple but useful for controlled experiments. The authors are also candid about limitations: no uncertainty information, no MHD knowledge, and they promise code and data (though the repo isn't public yet).

What's missing: error bars or multiple seeds; comparisons to published reconstruction models (Mai et al. and Wang et al. are cited but not used as baselines); out-of-device generalization tests; and an independent check on R. The per-dataset tuning of c1 also weakens the PILF claims.

If the authors reframed PILF as 'enforcing measurement consistency' rather than 'correcting inversion errors,' and added the missing evidence, the paper would be solid. As it stands, it's a conditional accept: serious referee, expected revisions. I'd bring it to reading group only if someone in the group is doing physics-informed ML for inverse problems.

Referee Report

4 major / 6 minor

Summary. The paper proposes a physics-informed deep learning architecture, 'Onion', for reconstructing 2D plasma emissivity profiles from line-integral soft X-ray diagnostic measurements. The architecture feeds a positional-encoded response-matrix 'physical information' (PI) side chain into a backbone network (VGG or ResNet) via element-wise multiplication, and augments the standard supervised loss with a line-integral consistency loss (loss2) that penalizes the difference between the back-projection of the network output and the measured chords. The authors evaluate the contributions of PI, a Softplus activation in the final fully connected layers, and the physics-informed loss function (PILF) on two synthetic datasets (noise-free, generated by their own forward model) and two experimental datasets from EAST and HL-2A, reporting reductions in the average relative reconstruction error E1 and the back-projection error E2. The paper also provides execution-time comparisons showing that the surrogate models are orders of magnitude faster than the F-B and NSGPT inversion codes.

Significance. If the claims were fully validated, the Onion architecture would be a useful, backbone-agnostic wrapper for embedding diagnostic response matrices into neural-network reconstructions, with potential impact on real-time plasma control and between-shot analysis. The paper contributes a synthetic data generator, two experimental datasets, and a clear description of the architecture and loss design, as well as a code/data availability commitment. The strongest asset is the architectural idea of multiplying learned features by fixed physical-information maps, which is simple and plausibly transferable. However, the current evidence is not sufficient to support the central claim that PILF corrects inversion-algorithm errors: the E2 improvement is largely a consequence of minimizing the same mismatch that E2 measures, and the paper lacks ground-truth validation, uncertainty quantification, multiple-seed statistics, and an independent calibration of the forward model. The significance is therefore conditional on additional validation.

major comments (4)

[Section 4.3, Eq. (3-5) and Eq. (4-2)] The reported reduction in E2 after adding PILF is largely by construction, because loss2 in Eq. (3-5) penalizes the same back-projection mismatch that E2 in Eq. (4-2) measures. Consequently, the abstract's claim that PILF 'reduces the errors associated with inversion algorithms' is not supported by E2 alone. Table 5 shows that the E2 improvement is accompanied by a large degradation in E1, e.g., on Exp_HL-2A E1 rises from 0.21e-2 to 1.94e-2 for VggOnion_PI and from 0.21e-2 to 1.93e-2 for ResOnion_PI. Without a known ground truth (or an independent diagnostic) demonstrating that the reduced E2 corresponds to a more accurate plasma profile, the interpretation that PILF corrects inversion errors remains unsupported; please provide validation on noisy synthetic data where the true profile is known, or temper the claim to consistency with the forward model.
[Section 2.1-2.2 and Section 3.2, Eq. (2-1)] The entire physical-consistency mechanism rests on the accuracy of the forward model x_i = R_i * y + Delta_i (Eq. 2-1). The paper provides no independent calibration or validation of the response matrices R_i against a known source, and the experimental datasets contain no uncertainty information. If the response matrices are miscalibrated or the systematic error Delta_i is not negligible, both the PI side chain and loss2 will actively pull predictions toward an incorrect measurement model. The authors should include a sensitivity analysis to perturbations in R (or a calibration check against an emissivity standard) and should state explicitly how the absence of Delta_i uncertainty affects the strength of the physical constraint.
[Section 4, Tables 3-5] All reported metrics come from a single training run per configuration, with no error bars, no repeated-seed statistics, and no statistical significance tests. Several of the claimed improvements are very small in absolute terms, such as the average E1 reduction of 0.06e-2 on experimental datasets from adding PI (Table 3), which may be within run-to-run variation. Please report mean and standard deviation over at least five independent training runs, and where possible compare with the baseline inversion algorithms (F-B code and NSGPT) on the same held-out test samples rather than only through execution time and aggregate dataset errors.
[Section 4.3, Table 5] The hyperparameter c1 that weights loss2 is set per dataset and per model (0.618 for Exp_EAST, 1.0 for Exp_HL-2A) with no sensitivity analysis, no principled selection criterion, and no discussion of how sensitive the E1/E2 trade-off is to this value. Since c1 directly controls the balance between matching the inversion-algorithm target and matching the raw measurements, the paper should report results across a range of c1 values and justify the chosen values; otherwise the strong E2 reduction on Exp_HL-2A (from 5.38e-2 to 0.76e-2) could simply reflect an overly large c1 that sacrifices profile fidelity for measurement consistency.

minor comments (6)

[Section 3.2, Eq. (3-1) to (3-3)] The theoretical motivation for multiplication over addition or concatenation relies on a single-neuron example and a statement that 'empirical testing of both approaches revealed suboptimal model performance'; please provide or cite the actual comparison results, otherwise the claim is not verifiable.
[Table 2] The execution-time comparison is not controlled: F-B code timings are cited from a laptop with an i7-8750H, NSGPT from an A100 server, while the surrogate models are timed on a different desktop with an RTX 3050; the reported speedup factors therefore mix hardware and method differences.
[Equation (3-4) vs. Eq. (4-1)] The training loss loss1 is a mean squared error while the evaluation metric E1 is a mean absolute relative error; the paper should state explicitly why the training objective and the evaluation metric differ, and whether an alternative loss (e.g., relative L1) was explored.
[Section 2.1] The paper correctly acknowledges that the experimental datasets lack uncertainty information, but the statement that this omission 'precludes the conduct of uncertainty-related research' should be reconciled with the strong claims about improving reconstruction accuracy, since the same lack of uncertainty also limits the conclusions that can be drawn about experimental performance.
[Section 2.2 and Section 4.1] The synthetic datasets consist of simple circular emissivity profiles, which are not representative of the peaked, hollow, or MHD-perturbed profiles found in real tokamak discharges; the transferability of the synthetic-data results to experimental conditions is therefore limited and should be stated as a caveat.
[Figure 19 and Section 4.3] The description of the PILF result on Exp_EAST says the in-out asymmetry 'seems to be in the opposite direction with respect to the target profile'; this qualitative statement should be either quantified (e.g., through a line-out comparison) or softened, since it is central to the claim that PILF corrects a core-region error.

Circularity Check

1 steps flagged · score 6.0 of 10

PILF's E2 improvement is the training objective itself; the claim that this corrects inversion-algorithm errors is not independently established.

fitted input called prediction [Abstract; Section 4.3; Eqs. (3-5), (4-2), Table 5]
"This section adopts experimental datasets for model training to account for these complexities and only 𝐸2 is considered to assess the model. … The incorporation of the physics-informed loss function has been shown to correct the model's predictions, bringing the back-projections closer to the actual inputs and reducing the errors associated with inversion algorithms."

loss2 (Eq. 3-5) is the squared back-projection mismatch (R_j·Net(x)-x_j)^2, and E2 (Eq. 4-2) is the same mismatch averaged in absolute form. Minimizing loss2 therefore reduces E2 by construction, so the Table 5 E2 drop is a training-objective decrease, not independent evidence that inversion-algorithm errors are corrected. On experimental data no ground-truth profile exists; Table 5 simultaneously shows E1 worsening (Exp_HL-2A VggOnion_PI: 0.21e-2 → 1.94e-2), so the only established effect is consistency with the assumed forward model. The interpretation that target profiles 'indeed contain errors' and that PILF corrects them is assumed, not derived.

full rationale

The paper's main architectural contributions—the PI multiplication side chain and the Softplus output activation—are tested through E1 against fixed target labels; those comparisons are self-contained and no circularity was found there. The circularity is confined to the PILF evaluation in Section 4.3. loss2 (Eq. 3-5) is the squared residual between the forward projection of the network output and the measured chords, and E2 (Eq. 4-2) is the same residual in mean-absolute form. After training with PILF, Table 5 reports a decrease in E2, but this is a decrease of the very objective being optimized; it cannot, by itself, show that 'errors associated with inversion algorithms' are reduced. The experimental datasets contain no independent ground-truth profile, and Table 5 shows E1 rising substantially (e.g., Exp_HL-2A VggOnion_PI from 0.21e-2 to 1.94e-2), so the model is moving away from the inversion-algorithm labels. The paper's claim that this movement corrects inversion errors rests on the unverified assumption that the labels are wrong in exactly the direction predicted by loss2 and that the response matrices R are perfectly calibrated. The paper's own disclaimers about missing uncertainty information and limited MHD physics are acknowledged correctness risks rather than circularities. Thus the score reflects one partially circular step: a training-objective decrease is presented as a physics correction.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central claims rest on assumptions about the forward model and label quality. The only tuned numerical parameter that directly affects the headline PILF result is c1, chosen per dataset. There are no invented physical entities.

free parameters (1)

c1 (weighting of loss2 in PILF) = 0.618 for Exp_EAST, 1.0 for Exp_HL-2A
The hyperparameter c1 controls the emphasis of the line-integral consistency loss (Eq. 3-8). It is manually selected per dataset in Section 4.3, and the reported E2 reductions depend on it.

assumptions (4)

domain assumption Line-integral forward model with known response matrices
Eq. 2-1 defines the measurement as R_i*y+Delta_i; the PILF (Eq. 3-5) and PI encoding assume this model and the response matrices are accurate.
domain assumption Inversion-algorithm outputs provide valid training labels
Target profiles y from F-B code (EAST) and NSGPT (HL-2A) are used as supervision; the paper notes they have mean relative error about 5e-2, but still treats closeness to them as reconstruction quality (E1).
standard math Universal approximation and trainability of chosen backbones
The empirical claim relies on the ability of VGG and ResNet to learn the mapping with the given architecture and data.
ad hoc to paper Multiplication fusion of PI features is beneficial
The motivation is a single-neuron linear example (Eq. 3-3) plus empirical testing, not a general theorem; the design is specific to this paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Physics-Informed Deep Learning Model for Line-integral Diagnostics Across Fusion Devices." pith.science (2026). https://pith.science/paper/ZVBNDBLQ

@misc{pith2026241200087,
  author       = {Pith},
  title        = {Pith review of: Physics-Informed Deep Learning Model for Line-integral Diagnostics Across Fusion Devices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZVBNDBLQ}},
  note         = {Machine review of arXiv:2412.00087}
}

read the original abstract

Rapid reconstruction of 2D plasma profiles from line-integral measurements is important in nuclear fusion. This paper introduces a physics-informed model architecture called Onion, that can enhance the performance of models and be adapted to various backbone networks. The model under Onion incorporates physical information by a multiplication process and applies the physics-informed loss function according to the principle of line integration. Prediction results demonstrate that the additional input of physical information improves the deep learning model's ability, leading to a reduction in the average relative error E_1 between the reconstruction profiles and the target profiles by approximately 0.84x10^(-2) on synthetic datasets and about 0.06x10^(-2) on experimental datasets. Furthermore, the implementation of the Softplus activation function in the final two fully connected layers improves model performance. This enhancement results in a reduction in the E_1 by approximately 1.06x10^(-2) on synthetic datasets and about 0.11x10^(-2) on experimental datasets. The incorporation of the physics-informed loss function has been shown to correct the model's predictions, bringing the back-projections closer to the actual inputs and reducing the errors associated with inversion algorithms. Besides, we have developed a synthetic data model to generate customized line-integral diagnostic datasets and have also collected soft x-ray diagnostic datasets from EAST and HL-2A. This study achieves reductions in reconstruction errors, and accelerates the development of surrogate models in fusion research.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 33 canonical work pages

[1]

Seo, J. et al. Development of an operation trajectory design algorithm for control of multiple 0D parameters using deep reinforcement learning in KSTAR. Nucl. Fusion 62, 086049 (2022)

work page 2022
[2]

Seo, J. et al. Multimodal Prediction of Tearing Instabilities in a Tokamak. in 2023 International Joint Conference on Neural Networks (IJCNN) 1–8 (IEEE, Gold Coast, Australia, 2023). doi:10.1109/IJCNN54540.2023.10191359

arXiv 2023
[3]

Vega, J. et al. Disruption prediction with artificial intelligence techniques in tokamak plasmas. Nat. Phys. 18, 741–750 (2022)

work page 2022
[4]

Fu, Y . et al. Machine learning control for disrup tion and tearing mode avoidance. Physics of Plasmas 27, 022501 (2020)

work page 2020
[5]

Liu, Z. Y . et al. Prediction of fishbone linear instability in tokamaks with machine learning methods. Nucl. Fusion 65, 016007 (2025)

work page 2025
[6]

& Mayo-García, R

Bustos, A., Ascasíbar, E., Cappa, A. & Mayo-García, R. Automatic identification of MHD modes in magnetic fluctuation spectrograms using deep learning techniques. Plasma Phys. Control. Fusion 63, 095001 (2021)

work page 2021
[7]

Kaptanoglu, A. A. et al. Exploring data-driven models for spatiotemporally local classification of Alfvén eigenmodes. Nucl. Fusion 62, 106014 (2022)

work page 2022
[8]

P., Hansen, C., Mauel, M

Wei, Y ., Levesque, J. P., Hansen, C., Mauel, M. E. & Navratil, G. A. MHD mode tracking using high -speed cameras and deep learning. Plasma Phys. Co ntrol. Fusion 65, 074002 (2023)

work page 2023

Show all 43 references

[9]

E., Seo, P

Lee, J. E., Seo, P. H., Bak, J. G. & Yun, G. S. A machine learning approach to identify the universality of solitary perturbations accompanying boundary bursts in magnetized toroidal plasmas. Sci Rep 11, 3662 (2021)

2021
[10]

Han, W. et al. Estimating cross-field particle transport at the outer midplane of TCV by tracking filaments with machine learning. Nucl. Fusion 63, 076025 (2023)

2023
[11]

Yang, K. N. et al. Neural network identification of the weakly coherent mode in I - mode discharge on EAST. Nucl. Fusion 64, 016035 (2024)

2024
[12]

W., Sabbagh, S

Piccione, A., Berkery, J. W., Sabbagh, S. A. & Andreopoulos, Y . Physics -guided machine learning approaches to predict the ideal stability properties of fusion plasmas. Nucl. Fusion 60, 046033 (2020)

2020
[13]

& Turnbull, A

Liu, Y ., Lao, L., Li, L. & Turnbull, A. D. Neural network based prediction of no-wall βN limits due to ideal external kink instabilities. Plasma Phys. Control. Fusion 62, 045001 (2020)

2020
[14]

Dong et al

G. Dong et al. Deep learning based surrogate models for first -principles global simulations of fusion plasmas. Nucl. Fusion 61, 126061 (2021)

2021
[15]

& Wang, Z

Li, H., Fu, Y ., Li, J. & Wang, Z. Machine learning of turbulent transport in fusion plasmas with neural network. Plasma Sci. Technol. 23, 115102 (2021)

2021
[16]

Van Mulders, S. et al. Rapid optimization of stationary tokamak plasmas in RAPTOR: demonstration for the ITER hybrid scenario with neural network surrogate transport model QLKNN. Nucl. Fusion 61, 086019 (2021)

2021
[17]

D., Logan, N

Clement, M. D., Logan, N. C. & Boyer, M. D. Neoclassical toroidal viscosity torque prediction via deep learning. Nucl. Fusion 62, 026022 (2022)

2022
[18]

& Tang, W

Kates-Harbeck, J., Svyatkovskiy, A. & Tang, W. Predicting disruptive instabilities in controlled fusion plasmas through deep learning. Nature 568, 526–531 (2019)

2019
[19]

& Karniadakis, G

Raissi, M., Perdikaris, P. & Karniadakis, G. E. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics 378, 686–707 (2019)

2019
[20]

Jalili, D. et al. Physics-informed neural networks for heat transfer prediction in two- phase flows. International Journal of Heat and Mass Transfer 221, 125089 (2024)

2024
[21]

& Sun, W

Liu, M., Liang, L. & Sun, W. A generic physics -informed neural network -based constitutive model for soft biological tissues. Computer Methods in Applied Mechanics and Engineering 372, 113402 (2020)

2020
[22]

& Wriggers, P

Wessels, H., Weißenfels, C. & Wriggers, P. The neural particle method – An updated Lagrangian physics informed neural network for c omputational fluid dynamics. Computer Methods in Applied Mechanics and Engineering 368, 113127 (2020)

2020
[23]

Sahli Costabal, F., Yang, Y ., Perdikaris, P., Hurtado, D. E. & Kuhl, E. Physics - Informed Neural Networks for Cardiac Activation Mapping. Front. Phys. 8, 42 (2020)

2020
[24]

Jagtap, A

Ameya D. Jagtap, A. D. J. & George Em Karniadakis, G. E. K. Extended Physics - Informed Neural Networks (XPINNs): A Generalized Space -Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations. CICP 28, 2002– 2041 (2020)

2020
[25]

Wang, Z. et al. Deep Learning Based Surrogate Model a fast Soft X -ray (SXR) Tomography on HL-2 a Tokamak. J Fusion Energ 43, 52 (2024)

2024
[26]

Mai, C. et al. Application of deep learning to soft x-ray tomography at EAST. Plasma Phys. Control. Fusion 64, 115009 (2022)

2022
[27]

Ferreira, D. R. Applications of Deep Learning to Nuclear Fusion Research. Preprint at http://arxiv.org/abs/1811.00333 (2018)

2018 arXiv
[28]

Tomography of m =1 mode structure in tokamak plasma using least - square-fitting method and Fourier –Bessel expansions

Nagayama, Y . Tomography of m =1 mode structure in tokamak plasma using least - square-fitting method and Fourier –Bessel expansions. Journal of Applied Physics 62, 2702– 2706 (1987)

1987
[29]

Chen, K. et al. 2-D soft x-ray arrays in the EAST. Review of Scientific Instruments 87, 063504 (2016)

2016
[30]

Neural network regression approaches to reconstruct properties of magnetic configuration from Wendelstein 7- X modeled heat load patterns

Blatzheim, M., Böckenhoff, D., & the Wendelstei n 7 -X Team. Neural network regression approaches to reconstruct properties of magnetic configuration from Wendelstein 7- X modeled heat load patterns. Nucl. Fusion 59, 126029 (2019)

2019
[31]

& Jenko, F

Matos, F., Svensson, J., Pavone, A., Odstrčil, T. & Jenko, F. Deep lea rning for Gaussian process soft x -ray tomography model selection in the ASDEX Upgrade tokamak. Review of Scientific Instruments 91, 103501 (2020)

2020
[32]

& Zisserman, A

Simonyan, K. & Zisserman, A. Very Deep Convolutional Networks for Large-Scale Image Recognition. Preprint at https://doi.org/10.48550/arXiv.1409.1556 (2015)

2015 doi arXiv
[33]

Reconstruction of soft X -ray and tungsten concentration profiles in Tokamaks using Bayesian method

Tianbo Wang. Reconstruction of soft X -ray and tungsten concentration profiles in Tokamaks using Bayesian method. (Ghent University, 2019)

2019
[34]

& Sun, J

He, K., Zhang, X., Ren, S. & Sun, J. Deep Residual Learning for Image Recognition. Preprint at http://arxiv.org/abs/1512.03385 (2015)

2015 arXiv
[35]

Vaswani, A. et al. Attention Is All You Need. Preprint at http://arxiv.org/abs/1706.03762 (2023)

2023 arXiv
[36]

Dosovitskiy, A. et al. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. Preprint at http://arxiv.org/abs/2010.11929 (2021)

2021 arXiv
[37]

Li, D. et al. Bayesian tomography and integrated data analysis in fusion diagnostics. Rev. Sci. Instrum. 87, 11E319 (2016)

2016
[38]

Kirkpatrick, S., Gelatt, C. D. & Vecchi, M. P. Optimization by Simulated Annealing. Science 220, 671–680 (1983)

1983
[39]

& Hutter, F

Loshchilov, I. & Hutter, F. SGDR: Stochastic Gradient Descent with Warm Restarts. Preprint at https://doi.org/10.48550/arXiv.1608.03983 (2017)

2017 doi arXiv
[40]

Smith, L. N. & Topin, N. Super -Convergence: Very Fast Training of Neural Networks Using Large Learning Rates. Preprint at https://doi.org/10.48550/arXiv.1708.07120 (2018)

2018 doi arXiv
[41]

Wu, T. et al. Compositional Generative Inverse Design. Preprint at http://arxiv.org/abs/2401.13171 (2024)

2024 arXiv
[42]

Liu, Z. et al. Plasma electron density profile tomography for EAST based on integrated data analysis. Nucl. Fusion 64, 126006 (2024)

2024
[43]

Kingma, D. P. & Ba, J. Adam: A Method for Stochastic Optimization. Preprint at https://doi.org/10.48550/arXiv.1412.6980 (2017). Appendix A. Adam Optimizer and Cosine Annealing Learning Rate Scheduler In this work, we employ a combination of the Adam optimizer and a cosine anne...

2017 doi arXiv

Pith tools

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

[1] [1]

Seo, J. et al. Development of an operation trajectory design algorithm for control of multiple 0D parameters using deep reinforcement learning in KSTAR. Nucl. Fusion 62, 086049 (2022)

work page 2022

[2] [2]

Seo, J. et al. Multimodal Prediction of Tearing Instabilities in a Tokamak. in 2023 International Joint Conference on Neural Networks (IJCNN) 1–8 (IEEE, Gold Coast, Australia, 2023). doi:10.1109/IJCNN54540.2023.10191359

arXiv 2023

[3] [3]

Vega, J. et al. Disruption prediction with artificial intelligence techniques in tokamak plasmas. Nat. Phys. 18, 741–750 (2022)

work page 2022

[4] [4]

Fu, Y . et al. Machine learning control for disrup tion and tearing mode avoidance. Physics of Plasmas 27, 022501 (2020)

work page 2020

[5] [5]

Liu, Z. Y . et al. Prediction of fishbone linear instability in tokamaks with machine learning methods. Nucl. Fusion 65, 016007 (2025)

work page 2025

[6] [6]

& Mayo-García, R

Bustos, A., Ascasíbar, E., Cappa, A. & Mayo-García, R. Automatic identification of MHD modes in magnetic fluctuation spectrograms using deep learning techniques. Plasma Phys. Control. Fusion 63, 095001 (2021)

work page 2021

[7] [7]

Kaptanoglu, A. A. et al. Exploring data-driven models for spatiotemporally local classification of Alfvén eigenmodes. Nucl. Fusion 62, 106014 (2022)

work page 2022

[8] [8]

P., Hansen, C., Mauel, M

Wei, Y ., Levesque, J. P., Hansen, C., Mauel, M. E. & Navratil, G. A. MHD mode tracking using high -speed cameras and deep learning. Plasma Phys. Co ntrol. Fusion 65, 074002 (2023)

work page 2023