Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

From Proxies to Fields: Spatiotemporal Reconstruction of Global Radiation from Sparse Sensor Sequences

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 A spatiotemporal neural operator trained on simulated dose maps reconstructs global cosmic radiation fields from 12 neutron monitors with sub-0.1% error and a 58,000x speedup.

desk verdict Useful surrogate for EXPACS, but the headline accuracy claim is contradicted by its own Table 1, and the framing overstates what is being reconstructed. read the letter →

arxiv 2506.12045 v1 pith:RNN36IC2 submitted 2025-05-24 cs.LG cs.AIeess.SP

Kazuma Kobayashi , Samrendra Roy , Seid Koric , Diab Abueidda , Syed Bahauddin Alam This is my paper · ORCID
classification cs.LGcs.AIeess.SP
keywords neuraloperatorsspatiotemporalreconstructioncosmicradiationdoseneutronmonitorsinverseproblemsreal-timeinferencedeeplearningforsciencefieldmapping
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that a single trained network can reconstruct the continuous global field of ground-level cosmic-ray dose from nothing more than the recent multi-day counting histories of a dozen fixed neutron monitors, without running a physics simulation at inference time. The authors build TRON, a neural operator with a recurrent temporal encoder in its branch network, and train it on 22 years of daily simulated dose maps. They report relative $L_2$ errors below 0.1% across input windows of 7 to 90 days, an inference time under 3 milliseconds per global map, and a speedup of more than 58,000x over the Monte Carlo simulator that generated the training targets. If accurate, this makes real-time global radiation monitoring for aviation and spaceflight feasible with a very sparse sensor network. The central caveat, flagged by the paper itself, is that the targets are simulated fields rather than measured dose rates.

What carries the argument

The central object is TRON (Temporal Radiation Operator Network), a neural operator formed by grafting a recurrent temporal encoder onto the DeepONet architecture. In the single-branch version, the 12 station time series are concatenated into one $T \times S$ input and passed through a four-layer LSTM (or GRU); the final hidden state becomes the branch latent $b$, while a two-layer feedforward trunk maps each query coordinate to a spatial latent $t$. The predicted dose at a point is $X = \sum_i b_i t_i + \beta$, an elementwise-product fusion summed over the 128 hidden dimensions. This mechanism carries the argument because the trunk lets the same trained model output at any coordinate, the recurrent branch turns the sensor history into a memory of solar modulation, and the single-branch joint encoding preserves cross-station correlations that the paper shows multi-branch models fragment.

What would settle it

Train TRON on inputs from only 11 of the 12 stations and compare its reconstructed dose field at the withheld station's location, and also compare against independent dosimeter readings taken during a solar energetic particle event; if the error jumps well above the reported 0.1% or disagrees with the dosimeters, the claim that 12 stations determine the global field is falsified.

Watch

Extended reading notes

Core claim

TRON treats dose reconstruction as a spatiotemporal inverse operator problem: map a tensor of neutron-count histories from 12 stations to a function that assigns an effective dose rate to any geographic coordinate. The best-performing configuration joins all station histories into one input and encodes them with a four-layer LSTM; a separate trunk network encodes longitude and latitude, and the two latent vectors are multiplied elementwise and summed to give the dose. Against daily simulated dose fields at 65,341 global points over 8,400 days, the single-branch LSTM variant keeps relative $L_2$ error below 0.1% for all tested sequence lengths (7–90 days), with the paper reporting a greater than 58,000x speedup over the simulation and sub-3 ms global inference. The paper interprets this as evidence that joint temporal encoding across sensors captures the global coherence of solar modulation, and that a resolution-agnostic operator can extrapolate to arbitrary spatial queries without retraining.

Load-bearing premise

The load-bearing premise is that the simulator-generated dose fields used as training targets are faithful representations of true ground-level radiation, and that 12 neutron monitors carry enough independent information to pin down the whole globe; without that, sub-0.1% error against simulation does not imply real-world accuracy.

Editorial extensions

If this is right

  • Operational systems can generate daily global dose maps continuously from existing neutron-monitor feeds, with no waiting on Monte Carlo simulation, because inference costs a few milliseconds per field.
  • The same trained checkpoint can be asked for dose at any coordinates, including grids finer or shifted from the 1-degree training grid, since the trunk decodes arbitrary spatial queries.
  • Input histories of 7, 30, 60, or 90 days can be mixed at inference time without retraining, letting an operational system adapt to data availability.
  • Single-branch joint encoding outperforms per-sensor encoders, so future TRON deployments for correlated geophysical fields should use joint temporal encoding.

Reading between the lines

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

  • The 0.1% error is measured against simulator output whose solar modulation comes from neutron-monitor data; an independent check would be to compare TRON against aircraft-borne or other dosimetry, which the paper does not do.
  • A station-holdout experiment would quantify whether 12 monitors truly determine the global field; the paper does not report how accuracy degrades as stations are removed.
  • The architecture is stated to be domain-agnostic, so a direct testable extension is to train the same spatiotemporal operator on sparse air-quality or seismic records and compare against dense reference fields; this is not claimed by the paper.
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 TRON, a DeepONet-style neural operator whose branch network is an LSTM or GRU encoder, and applies it to reconstructing global sea-level effective dose fields from sequences of 12 neutron-monitor count rates. The model is trained on 22 years of daily EXPACS simulations and evaluated on a temporally held-out test year. The authors report single-branch LSTM variants achieving relative L2 errors around 0.1%, sub-3-ms inference, and large speedups relative to an EXPACS reference run, and they argue that TRON is a general spatiotemporal inverse operator not limited to radiation dosimetry.

Significance. If the quantitative claims are accurate, TRON is a useful fast surrogate for EXPACS-style nowcasting of global cosmic-ray dose fields, and the architecture—a recurrent branch fused with a coordinate-based trunk—is a reasonable design for sparse-sequence-to-field reconstruction. The paper has notable strengths: a long (22-year) dataset, an ablation of single- vs. multi-branch temporal encoders, four sequence lengths, multiple error metrics, and explicit runtime benchmarking. However, the current presentation contains internal factual contradictions and an overstatement of the reconstruction claim. The reported errors are against EXPACS labels whose solar-modulation index is derived from neutron-monitor data, the same family of data used as TRON inputs, so the sub-0.1% result does not by itself establish independent physical reconstruction. The contribution is therefore best assessed as a surrogate-modeling result that needs qualification and additional validation before the broader claims can be accepted.

major comments (4)
  1. [Section 2.3, Table 1] Table 1 reports S-LSTM relative L2 errors of 1.12e-1% at 30 days and 1.10e-1% at 60 days, both above 0.1%; this directly contradicts the abstract and Section 2.5, which state that errors are below 0.1% and that S-LSTM sustains sub-0.1% errors across all sequence lengths. The authors should either correct the headline claim or explicitly define a different reporting convention (for example, using a different aggregation than the per-day relative L2 shown in Table 1).
  2. [Section 2.3, Supplement Table 5] The abstract and Section 2.5 claim that TRON generalizes across sequence lengths from 7 to 90 days and adapts without retraining, but Section 2.3 states that each model was trained and validated on its respective sequence-transformed dataset, and Supplement Table 5 lists separate training/validation/test partition sizes for each sequence length. No experiment transfers a model trained on one sequence length to another length. This cross-length generalization claim is not supported by the reported experiments and should be removed or tested explicitly.
  3. [Section 4.1] The reference dose fields are produced by EXPACS, whose solar modulation parameter (W-index) is itself calculated from neutron-monitor data, as stated in Section 4.1. Because the TRON inputs are NMDB neutron-monitor sequences from the same monitoring network family, the low test errors may largely reflect learning EXPACS's internal low-dimensional interpolation from a co-derived index rather than an independent reconstruction of the radiation field. The paper should quantify this circularity by holding out monitors not used in the W-index derivation, comparing against independent measurements or other dose models, and reporting performance during solar energetic particle events separately.
  4. [Abstract, Sections 1 and 2.5, Table 2] The paper repeatedly describes the speedup as exceeding 58,000x over Monte Carlo-based estimators, but EXPACS is described in Sections 1 and 4.1 as an analytical model (PARMA), not a Monte Carlo simulation. In addition, Table 2 compares 61.16 seconds of EXPACS runtime on a CPU against roughly 1 millisecond of TRON inference on an A100 GPU, which mixes hardware and simulation types. The speedup claim should be reworded to an analytic-model runtime comparison and either matched to a common hardware setting or stated with the hardware caveat.
minor comments (5)
  1. [Supplementary 1] The sentence 'missing data points were filled with polynomial interpolated' contains a grammar error, and the order, window, and validation of the polynomial imputation are not described; because gaps in ATHN and TERA are filled, the imputation should be documented in enough detail to assess its effect on the sequence data.
  2. [Table 1] The M-LSTM entry at 60 days is printed as 1.77 x 100, which is ambiguous; it should be formatted as 1.77e0 or 1.77, and the table should be checked for consistency of scientific notation across all entries.
  3. [Figure 4] The error maps would benefit from explicit color bars and units, since the absolute-error panels are otherwise difficult to interpret quantitatively; a common scale across subfigures would make the spatial comparisons more transparent.
  4. [Data and code availability] The statement that data and code are available 'on reasonable request' is not a verifiable reproducibility guarantee; depositing the code and processed datasets in a public repository would strengthen the paper and allow independent checks of the reported errors.
  5. [Section 2.6] The runtime comparison reports TRON inference in milliseconds but does not state whether the reported time includes pre- or post-processing, batching details, or GPU warm-up; clarifying these specifics would make the speedup claim more precise.

Circularity Check

1 steps flagged · score 5.0 of 10

Benchmark circularity: EXPACS dose labels are generated from neutron-monitor data, the same information source as TRON's 12 input sequences, so the sub-0.1% error is not an independent reconstruction result.

  1. self definitional [Section 2.1 (data generation) and Section 4.1 (EXPACS setup)]
    "Neutron monitor data were obtained from the Neutron Monitor Database (NMDB) ... Reference effective dose rates were simulated using the EXPACS toolkit ... The simulation specified the actual date (year, month, and day) to capture solar activity variations, enabling the calculation of the solar modulation parameter (W-index) based on neutron monitor data."

    TRON's input is the sequence of NMDB neutron-monitor counts (Y in Eq. 7), and the EXPACS fields used as training/test targets are not independent ground truth: EXPACS sets its daily solar-modulation parameter (W-index) from neutron-monitor data (Sec. 4.1). Since W-index is the dominant global driver of the PARMA/EXPACS dose field, the target already encodes information from the same monitor family as the inputs. The reported relative-L2 error therefore largely measures TRON's ability to emulate EXPACS's internal interpolation over (W-index, coordinates), not its ability to reconstruct true ground-level radiation from 12 sparse stations.

full rationale

The only load-bearing circularity is in the evaluation target: EXPACS computes W-index from neutron-monitor data, and TRON inputs are neutron-monitor sequences from NMDB, so the label and input share an information source. This makes the <0.1% error a surrogate-fidelity measure rather than an independent reconstruction validation. The self-citations (Kobayashi and Alam 2024; Koric and Abueidda 2023) are not load-bearing here: the static DeepONet limitation is independently documented by Lu et al. 2021, and the operator-learning background does not force the paper's conclusions. No uniqueness theorem is imported, and no ansatz is smuggled in via citation. Separately, Table 1 contradicts the abstract/Section 2.5 claim of 'sub-0.1% across all sequence lengths' (S-LSTM is 1.12e-1% at 30 days and 1.10e-1% at 60 days); this is a factual inconsistency and a correctness risk, not itself a circularity. Because the central reconstruction claim is partly reduced to the shared W-index/neutron-monitor information, a moderate score of 5 is appropriate.

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

The central claim rests on the validity of EXPACS as ground truth, the sufficiency of 12 neutron monitors, the adequacy of the temporal split, and the restriction to sea level. All are domain assumptions that the paper does not independently verify. No new physical entities or free parameters are introduced; the model hyperparameters (e.g., 128 hidden units, 4 layers) are fixed implementation choices.

assumptions (4)
  • domain assumption EXPACS/PARMA provides accurate reference dose fields for training and evaluation.
    The paper treats EXPACS output as ground truth without validation against measured radiation or independent dosimetry models.
  • domain assumption The 12 selected neutron monitor stations contain sufficient information to determine the global sea-level dose field.
    No coverage analysis, ablation over station subsets, or information-theoretic argument is provided in the paper.
  • domain assumption Chronological splitting with the last 365 days as test prevents temporal leakage and supports generalization claims.
    The training data spans two solar cycles, and the one-year holdout may still be within the distribution of solar activity seen during training; long-term extrapolation is not demonstrated.
  • domain assumption Sea-level altitude is the only regime relevant for the claimed reconstruction.
    The paper fixes altitude at sea level, but the motivation includes aviation and space contexts where altitude dependence is central.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Proxies to Fields: Spatiotemporal Reconstruction of Global Radiation from Sparse Sensor Sequences." pith.science (2026). https://pith.science/paper/RNN36IC2

@misc{pith2026250612045,
  author       = {Pith},
  title        = {Pith review of: From Proxies to Fields: Spatiotemporal Reconstruction of Global Radiation from Sparse Sensor Sequences},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RNN36IC2}},
  note         = {Machine review of arXiv:2506.12045}
}
read the original abstract

Accurate reconstruction of latent environmental fields from sparse and indirect observations is a foundational challenge across scientific domains-from atmospheric science and geophysics to public health and aerospace safety. Traditional approaches rely on physics-based simulators or dense sensor networks, both constrained by high computational cost, latency, or limited spatial coverage. We present the Temporal Radiation Operator Network (TRON), a spatiotemporal neural operator architecture designed to infer continuous global scalar fields from sequences of sparse, non-uniform proxy measurements. Unlike recent forecasting models that operate on dense, gridded inputs to predict future states, TRON addresses a more ill-posed inverse problem: reconstructing the current global field from sparse, temporally evolving sensor sequences, without access to future observations or dense labels. Demonstrated on global cosmic radiation dose reconstruction, TRON is trained on 22 years of simulation data and generalizes across 65,341 spatial locations, 8,400 days, and sequence lengths from 7 to 90 days. It achieves sub-second inference with relative L2 errors below 0.1%, representing a >58,000X speedup over Monte Carlo-based estimators. Though evaluated in the context of cosmic radiation, TRON offers a domain-agnostic framework for scientific field reconstruction from sparse data, with applications in atmospheric modeling, geophysical hazard monitoring, and real-time environmental risk forecasting.

Figures

Figures reproduced from arXiv: 2506.12045 by the authors.

Figure 1
Figure 1. Schematic overview of the inverse dose estimation framework. (a) Primary cosmic rays interact with the [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Architecture of the Temporal Radiation Operator Network (TRON). (a) Single-branch TRON: joint encoding [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Probability density distributions of relative [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of global effective dose field predictions and spatial error maps for the 90-day input sequence [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Training time (in seconds) for different model architectures across sequence lengths. Error bars represent the [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Fig.6 [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 6
Figure 6. Figure 6: Illustration of branch network architectures in TRON and the final fusion mechanism with the trunk output. [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: Daily neutron count rates recorded from 12 monitoring stations used in this study. Missing data from ATHN [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Single vs. Multiple Branches in DeepONet and S-DeepONet: Network Architecture Follows Coupling in Multiphysics Systems

    cs.LG 2025-07 conditional novelty 6.0 of 10

    For coupled multiphysics problems, single-branch DeepONet and S-DeepONet architectures predict more accurately than multi-branch MIONet-style designs, while multi-branch designs remain better for uncoupled or single-p...

  2. Distribution-Free Uncertainty-Aware Virtual Sensing via Conformalized Neural Operators

    cs.LG 2025-07 conditional novelty 4.0 of 10

    CMCO attaches conformalized Monte Carlo dropout intervals to DeepONet field predictions, achieving high average coverage but severe undercoverage in some test samples.

Reference graph

Works this paper leans on

58 extracted references · 51 canonical work pages · cited by 2 Pith papers

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence skip FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTION or pop #1 'skip if FUNCTION new.block.checka empty 'skip 'new.block if FUNCTION new.block.checkb...

    work page

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  3. [3]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter edition editor eprint howpublished institution journal key month note number organization pages publisher school series title type url volume year label INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.all := #1 'mid.sentence := #2 'after.sentence ...

  4. [4]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  5. [5]

    Cancer risk from exposure to galactic cosmic rays: implications for space exploration by human beings

    Francis A Cucinotta and Marco Durante. Cancer risk from exposure to galactic cosmic rays: implications for space exploration by human beings. The lancet oncology, 7 0 (5): 0 431--435, 2006

    work page 2006

  6. [6]

    Heavy ion carcinogenesis and human space exploration

    Marco Durante and Francis A Cucinotta. Heavy ion carcinogenesis and human space exploration. Nature Reviews Cancer, 8 0 (6): 0 465--472, 2008

    work page 2008

  7. [7]

    Galactic cosmic radiation leads to cognitive impairment and increased a plaque accumulation in a mouse model of alzheimer’s disease

    Jonathan D Cherry, Bin Liu, Jeffrey L Frost, Cynthia A Lemere, Jacqueline P Williams, John A Olschowka, and M Kerry O’Banion. Galactic cosmic radiation leads to cognitive impairment and increased a plaque accumulation in a mouse model of alzheimer’s disease. PloS one, 7 0 (12): 0 e53275, 2012

    work page 2012

  8. [8]

    Could spaceflight-associated immune system weakening preclude the expansion of human presence beyond earth’s orbit? Journal of leukocyte biology, 86 0 (5): 0 1027--1038, 2009

    Nathan Gu \'e guinou, C \'e cile Huin-Schohn, Matthieu Bascove, Jean-Luc Bueb, Eric Tschirhart, Christine Legrand-Frossi, and Jean-Pol Frippiat. Could spaceflight-associated immune system weakening preclude the expansion of human presence beyond earth’s orbit? Journal of leukocyte biology, 86 0 (5): 0 1027--1038, 2009

    work page 2009

Show all 58 references
  1. [9]

    Health risks from exposure to low levels of ionizing radiation

    VII Beir. Health risks from exposure to low levels of ionizing radiation. The National Academies report in brief, 2005

    2005

  2. [10]

    Evaluation of the Linear-Nonthreshold Dose-Response Model for Ionizing Radiation

    National Council on Radiation Protection and Measurements . Evaluation of the Linear-Nonthreshold Dose-Response Model for Ionizing Radiation. Number 136 in NCRP Report. National Council on Radiation Protection and Measurements, Bethesda, MD, 2001. ISBN 0-929600-69-X

    2001

  3. [11]

    Hormesis: the dose-response revolution

    Edward J Calabrese and Linda A Baldwin. Hormesis: the dose-response revolution. Annual review of pharmacology and toxicology, 43 0 (1): 0 175--197, 2003

    2003

  4. [12]

    The adaptive response in radiobiology: evolving insights and implications

    Sheldon Wolff. The adaptive response in radiobiology: evolving insights and implications. Environmental health perspectives, 106 0 (suppl 1): 0 277--283, 1998

    1998

  5. [13]

    The cosmic ray nucleonic component: the invention and scientific uses of the neutron monitor: keynote lecture

    John A Simpson. The cosmic ray nucleonic component: the invention and scientific uses of the neutron monitor: keynote lecture. 2000

    2000

  6. [14]

    Cosmic ray induced ionization model crac: Crii: An extension to the upper atmosphere

    Ilya G Usoskin, Gennady A Kovaltsov, and Irina A Mironova. Cosmic ray induced ionization model crac: Crii: An extension to the upper atmosphere. Journal of Geophysical Research: Atmospheres, 115 0 (D10), 2010

    2010

  7. [15]

    Cari-7a: Development and validation

    Kyle Copeland. Cari-7a: Development and validation. Radiation Protection Dosimetry, 175 0 (4): 0 419--431, 2017

    2017

  8. [16]

    Analytical model for estimating terrestrial cosmic ray fluxes nearly anytime and anywhere in the world: Extension of parma/expacs

    Tatsuhiko Sato. Analytical model for estimating terrestrial cosmic ray fluxes nearly anytime and anywhere in the world: Extension of parma/expacs. PloS one, 10 0 (12): 0 e0144679, 2015

    2015

  9. [17]

    Analytical model for estimating the zenith angle dependence of terrestrial cosmic ray fluxes

    Tatsuhiko Sato. Analytical model for estimating the zenith angle dependence of terrestrial cosmic ray fluxes. PloS one, 11 0 (8): 0 e0160390, 2016

    2016

  10. [18]

    Deep learning-based effective fine-grained weather forecasting model

    Pradeep Hewage, Marcello Trovati, Ella Pereira, and Ardhendu Behera. Deep learning-based effective fine-grained weather forecasting model. Pattern Analysis and Applications, 24 0 (1): 0 343--366, 2021

    2021

  11. [19]

    Accurate medium-range global weather forecasting with 3d neural networks

    Kaifeng Bi, Lingxi Xie, Hengheng Zhang, Xin Chen, Xiaotao Gu, and Qi Tian. Accurate medium-range global weather forecasting with 3d neural networks. Nature, 619 0 (7970): 0 533--538, 2023

    2023

  12. [20]

    Deep learning to represent subgrid processes in climate models

    Stephan Rasp, Michael S Pritchard, and Pierre Gentine. Deep learning to represent subgrid processes in climate models. Proceedings of the national academy of sciences, 115 0 (39): 0 9684--9689, 2018

    2018

  13. [21]

    Pushing the frontiers in climate modelling and analysis with machine learning

    Veronika Eyring, William D Collins, Pierre Gentine, Elizabeth A Barnes, Marcelo Barreiro, Tom Beucler, Marc Bocquet, Christopher S Bretherton, Hannah M Christensen, Katherine Dagon, et al. Pushing the frontiers in climate modelling and analysis with machine learning. Nature Cl...

    2024

  14. [22]

    A survey of machine learning and deep learning in remote sensing of geological environment: Challenges, advances, and opportunities

    Wei Han, Xiaohan Zhang, Yi Wang, Lizhe Wang, Xiaohui Huang, Jun Li, Sheng Wang, Weitao Chen, Xianju Li, Ruyi Feng, et al. A survey of machine learning and deep learning in remote sensing of geological environment: Challenges, advances, and opportunities. ISPRS Journal of Photo...

    2023

  15. [23]

    Using artificial intelligence and data fusion for environmental monitoring: A review and future perspectives

    Yassine Himeur, Bhagawat Rimal, Abhishek Tiwary, and Abbes Amira. Using artificial intelligence and data fusion for environmental monitoring: A review and future perspectives. Information Fusion, 86: 0 44--75, 2022

    2022

  16. [24]

    Deep learning to infer eddy heat fluxes from sea surface height patterns of mesoscale turbulence

    Tom M George, Georgy E Manucharyan, and Andrew F Thompson. Deep learning to infer eddy heat fluxes from sea surface height patterns of mesoscale turbulence. Nature communications, 12 0 (1): 0 800, 2021

    2021

  17. [25]

    Spatial prediction of groundwater potential mapping based on convolutional neural network (cnn) and support vector regression (svr)

    Mahdi Panahi, Nitheshnirmal Sadhasivam, Hamid Reza Pourghasemi, Fatemeh Rezaie, and Saro Lee. Spatial prediction of groundwater potential mapping based on convolutional neural network (cnn) and support vector regression (svr). Journal of Hydrology, 588: 0 125033, 2020

    2020

  18. [26]

    A gt-lstm spatio-temporal approach for winter wheat yield prediction: From the field scale to county scale

    Enhui Cheng, Fumin Wang, Dailiang Peng, Bing Zhang, Bin Zhao, Wenjuan Zhang, Jinkang Hu, Zihang Lou, Songlin Yang, Hongchi Zhang, et al. A gt-lstm spatio-temporal approach for winter wheat yield prediction: From the field scale to county scale. IEEE Transactions on Geoscience ...

    2024

  19. [27]

    Checkpoint data-driven gcn-gru vehicle trajectory and traffic flow prediction

    Deyong Guan, Na Ren, Ke Wang, Qi Wang, and Hualong Zhang. Checkpoint data-driven gcn-gru vehicle trajectory and traffic flow prediction. Scientific Reports, 14 0 (1): 0 1--17, 2024

    2024

  20. [28]

    Neural operator: Graph kernel network for partial differential equations

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Neural operator: Graph kernel network for partial differential equations. arXiv preprint arXiv:2003.03485, 2020

    2003 arXiv

  21. [29]

    Fourier neural operator with learned deformations for pdes on general geometries

    Zongyi Li, Daniel Zhengyu Huang, Burigede Liu, and Anima Anandkumar. Fourier neural operator with learned deformations for pdes on general geometries. Journal of Machine Learning Research, 24 0 (388): 0 1--26, 2023

    2023

  22. [30]

    Learning nonlinear operators via deeponet based on the universal approximation theorem of operators

    Lu Lu, Pengzhan Jin, Guofei Pang, Zhongqiang Zhang, and George Em Karniadakis. Learning nonlinear operators via deeponet based on the universal approximation theorem of operators. Nature machine intelligence, 3 0 (3): 0 218--229, 2021

    2021

  23. [31]

    Data-driven and physics-informed deep learning operators for solution of heat conduction equation with parametric heat source

    Seid Koric and Diab W Abueidda. Data-driven and physics-informed deep learning operators for solution of heat conduction equation with parametric heat source. International Journal of Heat and Mass Transfer, 203: 0 123809, 2023

    2023

  24. [32]

    Sequential deep operator networks for predicting full-field solutions under time-dependent loads

    Junyan He, Shashank Kushwaha, Jaewan Park, Seid Koric, Diab Abueidda, and Iwona Jasiuk. Sequential deep operator networks for predicting full-field solutions under time-dependent loads. Engineering Applications of Artificial Intelligence, 127: 0 107258, 2024 a

    2024

  25. [33]

    Deep learning operator network for plastic deformation with variable loads and material properties

    Seid Koric, Asha Viswantah, Diab W Abueidda, Nahil A Sobh, and Kamran Khan. Deep learning operator network for plastic deformation with variable loads and material properties. Engineering with Computers, 40 0 (2): 0 917--929, 2024

    2024

  26. [34]

    Predictions of transient vector solution fields with sequential deep operator network

    Junyan He, Shashank Kushwaha, Jaewan Park, Seid Koric, Diab Abueidda, and Iwona Jasiuk. Predictions of transient vector solution fields with sequential deep operator network. Acta Mechanica, pages 1--16, 2024 b

    2024

  27. [35]

    Advanced deep operator networks to predict multiphysics solution fields in materials processing and additive manufacturing

    Shashank Kushwaha, Jaewan Park, Seid Koric, Junyan He, Iwona Jasiuk, and Diab Abueidda. Advanced deep operator networks to predict multiphysics solution fields in materials processing and additive manufacturing. Additive Manufacturing, 88: 0 104266, 2024

    2024

  28. [36]

    Skilful precipitation nowcasting using deep generative models of radar

    Suman Ravuri, Karel Lenc, Matthew Willson, Dmitry Kangin, Remi Lam, Piotr Mirowski, Megan Fitzsimons, Maria Athanassiadou, Sheleem Kashem, Sam Madge, et al. Skilful precipitation nowcasting using deep generative models of radar. Nature, 597 0 (7878): 0 672--677, 2021

    2021

  29. [37]

    o wer, James Lottes, Stephan Rasp, Peter D \

    Dmitrii Kochkov, Janni Yuval, Ian Langmore, Peter Norgaard, Jamie Smith, Griffin Mooers, Milan Kl \"o wer, James Lottes, Stephan Rasp, Peter D \"u ben, et al. Neural general circulation models for weather and climate. Nature, 632 0 (8027): 0 1060--1066, 2024

    2024

  30. [38]

    Deep learning for twelve hour precipitation forecasts

    Lasse Espeholt, Shreya Agrawal, Casper S nderby, Manoj Kumar, Jonathan Heek, Carla Bromberg, Cenk Gazen, Rob Carver, Marcin Andrychowicz, Jason Hickey, et al. Deep learning for twelve hour precipitation forecasts. Nature communications, 13 0 (1): 0 1--10, 2022

    2022

  31. [39]

    Fourcastnet: A global data-driven weather forecasting model using adaptive fourier neural operators

    Jaideep Pathak, Shashank Subramanian, Peter Harrington, and et al. Fourcastnet: A global data-driven weather forecasting model using adaptive fourier neural operators. Advances in Neural Information Processing Systems, 35: 0 28804--28818, 2022

    2022

  32. [40]

    Applications and usage of the real-time neutron monitor database

    H Mavromichalaki, Athanasios Papaioannou, Christina Plainaki, C Sarlanis, G Souvatzoglou, M Gerontidou, M Papailiou, E Eroshenko, A Belov, V Yanke, et al. Applications and usage of the real-time neutron monitor database. Advances in Space Research, 47 0 (12): 0 2210--2222, 2011

    2011

  33. [41]

    Nmdb: The neutron monitor database

    NMDB Database . Nmdb: The neutron monitor database. URL https://www.nmdb.eu/. Real-Time Database for high-resolution Neutron Monitor measurements. Accessed: 2024-12-03

    2024

  34. [42]

    Novel deeponet architecture to predict stresses in elastoplastic structures with variable complex geometries and loads

    Junyan He, Seid Koric, Shashank Kushwaha, Jaewan Park, Diab Abueidda, and Iwona Jasiuk. Novel deeponet architecture to predict stresses in elastoplastic structures with variable complex geometries and loads. Computer Methods in Applied Mechanics and Engineering, 415: 0 116277, 2023

    2023

  35. [43]

    Deep neural operator-driven real-time inference to enable digital twin solutions for nuclear energy systems

    Kazuma Kobayashi and Syed Bahauddin Alam. Deep neural operator-driven real-time inference to enable digital twin solutions for nuclear energy systems. Scientific reports, 14 0 (1): 0 2101, 2024

    2024

  36. [44]

    EXPACS : EX cel-based program for calculating atmospheric cosmic-ray spectrum

    Japan Atomic Energy Agency . EXPACS : EX cel-based program for calculating atmospheric cosmic-ray spectrum. http://phits.jaea.go.jp/expacs/. Accessed: November 4, 2024

    2024

  37. [45]

    u ckiger, M Gurtner, MR Moser, and R B \

    Laurent Desorgher, EO Fl \"u ckiger, M Gurtner, MR Moser, and R B \"u tikofer. Atmocosmics: A geant 4 code for computing the interaction of cosmic rays with the earth's atmosphere. International Journal of Modern Physics A, 20 0 (29): 0 6802--6804, 2005

    2005

  38. [46]

    Athens neutron monitor and its aspects in the cosmic-ray variations studies

    H Mavromichalaki, C Sarlanis, G Souvatzoglou, S Tatsis, A Belov, E Eroshenko, V Yanke, and A Pchelkin. Athens neutron monitor and its aspects in the cosmic-ray variations studies. Proceedings of the 27th International Cosmic Ray Conference, 10: 0 4099, 2001

    2001

  39. [47]

    The new athens center on data processing from the neutron monitor network in real time

    Mavromichalaki, Souvatzoglou, Sarlanis, Mariatos, Gerontidou, Papaioannou, Plainaki, Tatsis, Belov, Eroshenko, et al. The new athens center on data processing from the neutron monitor network in real time. Annales Geophysicae, 23 0 (9): 0 3103--3110, 2005

    2005

  40. [48]

    Souvatzoglou, H

    G. Souvatzoglou, H. Mavromichalaki, C. Sarlanis, G. Mariatos, A. Belov, E. Eroshenko, and V. Yanke. Real-time gle alert in the anmodap center for december 13, 2006. Advances in Space Research, 43 0 (4): 0 728--734, 2009. ISSN 0273-1177. doi:https://doi.org/10.1016/j.asr.2008.0...

    2006 doi

  41. [49]

    Flückiger and Rolf Bütikofer

    Erwin O. Flückiger and Rolf Bütikofer. Swiss neutron monitors and cosmic ray research at jungfraujoch. Advances in Space Research, 44 0 (10): 0 1155--1159, 2009. ISSN 0273-1177. doi:https://doi.org/10.1016/j.asr.2008.10.043. URL https://www.sciencedirect.com/science/article/pi...

    2009 doi

  42. [50]

    URL https://www.nmdb.eu/station/newk/

    Newark/swarthmore detail, 2021. URL https://www.nmdb.eu/station/newk/. Accessed on February 20, 2025. Information page for the Newark/Swarthmore neutron monitor station on the NMDB website

    2021

  43. [51]

    New cosmogenic nuclide constraints on late glacial and holocene glacier fluctuations in the sub-antarctic indian ocean (kerguelen islands, 49° s)

    Joanna Charton, Irene Schimmelpfennig, Vincent Jomelli, Guillaume Delpech, Pierre-Henri Blard, R \'e gis Braucher, Deborah Verfaillie, Vincent Favier, Vincent Rinterknecht, Hugues Goosse, et al. New cosmogenic nuclide constraints on late glacial and holocene glacier fluctuatio...

    2022

  44. [52]

    A quarter of a century of relativistic solar cosmic ray events recorded by the oulu neutron monitor

    H Kananen, PJ Tanskanen, LC Gentile, MA Shea, and DF Smart. A quarter of a century of relativistic solar cosmic ray events recorded by the oulu neutron monitor. international cosmic ray conference, 3: 0 145, 1991

    1991

  45. [53]

    Upgrade of apatity neutron monitor

    Yu Balabin, E Vashenyuk, B Gzdevsky, and A Germanenko. Upgrade of apatity neutron monitor. Journal of Physics: Conference Series, 632 0 (1): 0 012047, 2015

    2015

  46. [54]

    URL https://www.nmdb.eu/station/fsmt/

    Fort Smith Station Detail , 2021. URL https://www.nmdb.eu/station/fsmt/. Accessed on February 20, 2025. Information page for the FSMT neutron monitor station on the NMDB website

    2021

  47. [55]

    URL https://www.nmdb.eu/station/invk/

    Inuvik station detail, 2021. URL https://www.nmdb.eu/station/invk/. Accessed on February 20, 2025. Information page for the Inuvik neutron monitor station on the NMDB website

    2021

  48. [56]

    Neutron monitor temperature coefficients: Measurements for bf3 and 3he counter tubes

    P Evenson. Neutron monitor temperature coefficients: Measurements for bf3 and 3he counter tubes. 2005

    2005

  49. [57]

    Long-term decline of south pole neutron rates

    JW Bieber, J Clem, D Desilets, P Evenson, D Lal, C Lopate, and R Pyle. Long-term decline of south pole neutron rates. Journal of Geophysical Research: Space Physics, 112 0 (A12), 2007

    2007

  50. [58]

    URL https://www.nmdb.eu/station/tera/

    Terre adelie detail, 2021. URL https://www.nmdb.eu/station/tera/. Accessed on February 20, 2025. Information page for the Terre Adelie neutron monitor station on the NMDB website

    2021

Pith tools

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