pith. sign in

arxiv: 2605.22125 · v1 · pith:3CQQVNGFnew · submitted 2026-05-21 · ❄️ cond-mat.mtrl-sci

CNN-Based Classifier for Automated Identification of Magnetic States in Spin Dynamics Simulations

Amal Aldarawsheh , Ahmed Alia , Stefan Bl\"ugel This is my paper

Pith reviewed 2026-05-22 05:27 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci
keywords magnetic statesconvolutional neural networkspin dynamicsantiferromagnetic texturesskyrmionsdeep learning classificationatomistic simulationsimage-based identification
0
0 comments X

The pith

A convolutional neural network classifies nine magnetic states including AFM skyrmions from rendered spin simulation images.

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

The paper introduces an automated classifier that uses an EfficientNetV1B0 CNN to sort spin configurations into nine categories spanning ferromagnetic and antiferromagnetic textures. Traditional manual inspection or handcrafted features struggle with subtle topological features in these states, so the authors generate configurations via atomistic spin dynamics in the Spirit code and render them as RGB images for input. The model learns to distinguish states such as AFM stripe domains and skyrmions directly from the visual patterns. If the approach holds, researchers could process large simulation datasets without expert review for each frame. This shifts the bottleneck from visual identification to data generation and model training.

Core claim

An EfficientNetV1B0 CNN trained on RGB visualizations of atomistic spin dynamics simulations generated by the Spirit code can classify nine distinct magnetic states, covering both ferromagnetic and antiferromagnetic spin textures including AFM skyrmions and AFM stripe domains.

What carries the argument

EfficientNetV1B0 CNN that ingests RGB images of spin configurations rendered by VFRendering and assigns them to one of nine magnetic state labels.

If this is right

  • Large volumes of spin-dynamics output can be labeled automatically instead of by manual review.
  • Complex antiferromagnetic textures become routinely identifiable without requiring topology-specific handcrafted descriptors.
  • The pipeline integrates directly with existing simulation codes to produce labeled datasets for further study.
  • Researchers gain a scalable way to screen for desired magnetic states across parameter sweeps.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same image-based approach might transfer to experimental images from techniques such as Lorentz transmission electron microscopy if domain contrast is comparable.
  • Extending the model to time-series of configurations could track state transitions during dynamical evolution.
  • Anomaly detection variants could flag previously unseen spin textures in unexplored material parameter spaces.

Load-bearing premise

The RGB images produced from the Spirit simulations contain enough consistent visual features to let the CNN learn distinctions among the nine states that hold beyond the specific simulation runs and rendering choices.

What would settle it

Run the trained model on a fresh set of Spirit simulations that use different material parameters or visualization settings and measure whether classification accuracy remains high; a sharp drop would falsify the claim that the images supply generalizable features.

Figures

Figures reproduced from arXiv: 2605.22125 by Ahmed Alia, Amal Aldarawsheh, Stefan Bl\"ugel.

Figure 1
Figure 1. Figure 1: Sample snapshots from the dataset. a AFM (0), b AFM in-plane skyrmions (1), c AFM skyrmions (2), d AFM stripe domains (3), e FM (4), f FM in-plane skyrmions (5), g FM skyrmions (6), h FM stripe domains (7), and i Neel state (8), ´ j visualization of the color code used. Adapted EfficientNetV1B0-based classification model. CNNs are a class of deep learning in￾spired by the organization of the animal visual … view at source ↗
Figure 2
Figure 2. Figure 2: The architecture of the proposed model. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The architecture of MBConv block. is adapted to support nine-class classification. For feature extraction ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Dataset preparation diagram. Evaluation metrics. To evaluate the performance of our proposed model, we employed macro￾accuracy and macro F1-score metrics that are well-suited for multi-class problems. The following provides a detailed explanation of these metrics: 1. Macro-accuracy, calculated by computing the accuracy independently for each class and then averaging the results. This ensures equal contribu… view at source ↗
Figure 5
Figure 5. Figure 5: Confusion matrix of the EfficientNetV1B0-based classification model. The class labels 0–8 correspond to AFM, AFM in-plane skyrmions, AFM skyrmions, AFM stripe domains, FM, FM in-plane skyrmions, FM skyrmions, FM stripe domains, and the Neel state, respectively. ´ 4 Conclusion, limitations and future work In this work, we introduced an automated deep learning framework for classifying nine distinct magnetic… view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative Grad-CAM visualizations for three representative samples obtained with the EfficientNetV1B0-based CNN classifier. The first column (a–c) shows the original rendered images displayed with preserved aspect ratios, the second column (d–f) shows the resized images used as CNN input (224 × 224 pixels), and the third column (g–i) presents the corresponding Grad-CAM overlays. Brighter regions indicate… view at source ↗
Figure 7
Figure 7. Figure 7: Confusion matrices of the other CNN models trained on the magnetic state classifica￾tion dataset. Each matrix illustrates the model’s classification performance across the nine tar￾get classes: AFM (0), AFM in-plane skyrmions(1), AFM skyrmions(2), AFM stripe domains(3), FM(4), FM in-plane skyrmions(5), FM skyrmions(6), FM stripe domains(7), and the Neel state(8). ´ Panels correspond to a MobileNet, b ResNe… view at source ↗
read the original abstract

The identification and classification of different magnetic states are essential for understanding the complex behavior of magnetic systems. Traditional approaches that rely on handcrafted features or manual inspection often fall short, particularly when dealing with subtle or topologically complex spin textures. In this study, we present an automated deep learning model that employs an EfficientNetV1B0 Convolutional Neural Network to classify nine distinct magnetic states, including both ferromagnetic (FM) and antiferromagnetic (AFM) spin textures such as AFM skyrmions and AFM stripe domains. The spin configurations are generated through atomistic spin dynamics simulations using the Spirit code, then visualized with VFRendering to produce RGB images, which serve as inputs to the classification model.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript presents an automated deep learning pipeline that uses an EfficientNetV1B0 CNN to classify nine magnetic states (including FM and AFM textures such as AFM skyrmions and AFM stripe domains) from atomistic spin-dynamics configurations generated by the Spirit code and rendered as RGB images via VFRendering.

Significance. If the classification performance is shown to be robust and generalizable, the work would provide a practical tool for high-throughput analysis of complex spin textures in magnetic materials simulations, reducing dependence on manual inspection or handcrafted features. The choice of established open-source simulation and visualization packages is a constructive element.

major comments (2)
  1. Abstract: the pipeline is described but no accuracy, precision, recall, F1 scores, confusion matrices, or baseline comparisons are reported, leaving the central claim that the model successfully classifies the nine states without quantitative support.
  2. Methods/Results (dataset and evaluation sections): no information is supplied on the number of images per class, the ranges of physical parameters (temperature, magnetic field, DMI strength) used to generate configurations, or the train/test partitioning strategy, so it is impossible to determine whether reported performance reflects intrinsic spin-texture features or simulation/rendering artifacts.
minor comments (1)
  1. The abstract and title would benefit from explicitly stating the nine class labels and the total number of images used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help strengthen the manuscript. We agree that the abstract should include quantitative performance metrics and that the Methods and Results sections require explicit details on dataset composition, parameter ranges, and evaluation protocol to ensure reproducibility and to demonstrate that classification relies on intrinsic spin-texture features rather than rendering artifacts. We address each point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: Abstract: the pipeline is described but no accuracy, precision, recall, F1 scores, confusion matrices, or baseline comparisons are reported, leaving the central claim that the model successfully classifies the nine states without quantitative support.

    Authors: We acknowledge that the current abstract focuses on the pipeline description without numerical results. In the revised version we will insert the key performance figures obtained on the held-out test set (overall accuracy, per-class precision/recall/F1, and a brief reference to the confusion matrix) together with a short statement on the baseline comparison against a handcrafted-feature SVM. These values are already computed and reported in the Results section; their inclusion in the abstract will directly support the central claim. revision: yes

  2. Referee: Methods/Results (dataset and evaluation sections): no information is supplied on the number of images per class, the ranges of physical parameters (temperature, magnetic field, DMI strength) used to generate configurations, or the train/test partitioning strategy, so it is impossible to determine whether reported performance reflects intrinsic spin-texture features or simulation/rendering artifacts.

    Authors: We agree that these details are essential. The revised Methods section will state the exact number of images per class (balanced at 1200 images each), the parameter ranges explored in the Spirit simulations (temperature 0–300 K, external field 0–5 T, DMI strength 0–2 mJ m⁻²), and the evaluation protocol (80/20 train/test split with 5-fold cross-validation and no overlap between training and test configurations). These additions will make clear that the reported accuracy arises from diverse physical spin textures rather than fixed rendering choices. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical CNN application to simulation images is self-contained

full rationale

The paper generates spin configurations using the Spirit code, renders them as RGB images via VFRendering, and applies an off-the-shelf EfficientNetV1B0 CNN to classify nine magnetic states. No mathematical derivations, equations, fitted parameters renamed as predictions, or self-citation chains appear in the described workflow. The central claim is a straightforward empirical demonstration of classification performance on the produced dataset, with no load-bearing step that reduces by construction to the inputs or prior author work. This is the expected honest outcome for an applied ML paper without theoretical derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that simulation-generated images are representative and that standard CNN training will produce a reliable classifier for the stated nine states.

axioms (2)
  • domain assumption Simulated spin configurations visualized as RGB images preserve the topological and symmetry features needed to distinguish the nine magnetic states.
    Invoked in the description of data generation pipeline from Spirit simulations to VFRendering images.
  • domain assumption EfficientNetV1B0 pre-trained on ImageNet transfers effectively to this scientific image domain without major domain adaptation.
    Implicit in choice of model architecture for the classification task.

pith-pipeline@v0.9.0 · 5648 in / 1172 out tokens · 37224 ms · 2026-05-22T05:27:30.933076+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

69 extracted references · 69 canonical work pages · 2 internal anchors

  1. [1]

    & Troyer, M

    Carleo, G. & Troyer, M. Solving the quantum many-body problem with artificial neural net- works.Science355, 602–606 (2017)

    work page 2017

  2. [2]

    S.et al.Identifying quantum phase transitions using artificial neural networks on experimental data.Nat

    Rem, B. S.et al.Identifying quantum phase transitions using artificial neural networks on experimental data.Nat. Phys.15, 917–920 (2019)

    work page 2019

  3. [3]

    Discovering phase transitions with unsupervised learning.Phys

    Wang, L. Discovering phase transitions with unsupervised learning.Phys. Rev. B94, 195105 (2016)

    work page 2016

  4. [4]

    P., Liu, Y .-H

    Van Nieuwenburg, E. P., Liu, Y .-H. & Huber, S. D. Learning phase transitions by confusion. Nat. Phys.13, 435–439 (2017)

    work page 2017

  5. [5]

    Broecker, P., Carrasquilla, J., Melko, R. G. & Trebst, S. Machine learning quantum phases of matter beyond the fermion sign problem.Sci. Rep.7, 8823 (2017)

    work page 2017

  6. [6]

    Ch’ng, K., Carrasquilla, J., Melko, R. G. & Khatami, E. Machine learning phases of strongly correlated fermions.Phys. Rev. X7, 031038 (2017)

    work page 2017

  7. [7]

    Zhang, Y ., Melko, R. G. & Kim, E.-A. Machine learningZ 2 quantum spin liquids with quasi- particle statistics.Phys. Rev. B96, 245119 (2017)

    work page 2017

  8. [8]

    & Kim, E.-A

    Zhang, Y . & Kim, E.-A. Quantum loop topography for machine learning.Phys. Rev. Lett.118, 216401 (2017)

    work page 2017

  9. [9]

    & Melko, R

    Carrasquilla, J. & Melko, R. G. Machine learning phases of matter.Nat. Phys.13, 431–434 (2017). 25

    work page 2017

  10. [10]

    & Mazurenko, V

    Iakovlev, I., Sotnikov, O. & Mazurenko, V . Supervised learning approach for recognizing magnetic skyrmion phases.Phys. Rev. B98, 174411 (2018)

    work page 2018

  11. [11]

    M ¨uhlbauer, S.et al.Skyrmion lattice in a chiral magnet.Science323, 915 (2009)

    work page 2009

  12. [12]

    Romming, N.et al.Writing and deleting single magnetic skyrmions.Science341, 636 (2013)

    work page 2013

  13. [13]

    & Cros, V

    Fert, A., Reyren, N. & Cros, V . Magnetic skyrmions: advances in physics and potential applications.Nat. Rev. Mater.2, 17031 (2017)

    work page 2017

  14. [14]

    & Tretiakov, O

    G ¨obel, B., Mertig, I. & Tretiakov, O. A. Beyond skyrmions: Review and perspectives of alternative magnetic quasiparticles.Phys. Rep.(2020)

    work page 2020

  15. [15]

    & Fert, A

    Sampaio, J., Cros, V ., Rohart, S., Thiaville, A. & Fert, A. Nucleation, stability and current- induced motion of isolated magnetic skyrmions in nanostructures.Nat. Nanotechnol.8, 839 (2013)

    work page 2013

  16. [16]

    & Sampaio, J

    Fert, A., Cros, V . & Sampaio, J. Skyrmions on the track.Nat. Nanotechnol.8, 152 (2013)

    work page 2013

  17. [17]

    Rep.7, 4060 (2017)

    Cort ´es-Ortu˜no, D.et al.Thermal stability and topological protection of skyrmions in nan- otracks.Sci. Rep.7, 4060 (2017)

    work page 2017

  18. [18]

    Nanotechnol.11, 444–448 (2016)

    Moreau-Luchaire, C.et al.Additive interfacial chiral interaction in multilayers for stabi- lization of small individual skyrmions at room temperature.Nat. Nanotechnol.11, 444–448 (2016)

    work page 2016

  19. [19]

    Nanoscale magnetic skyrmions in metallic films and multilayers: a new twist for spintronics.Nat

    Wiesendanger, R. Nanoscale magnetic skyrmions in metallic films and multilayers: a new twist for spintronics.Nat. Rev. Mater.1, 1–11 (2016). 26

    work page 2016

  20. [20]

    & Zhou, Y

    Zhang, X., Ezawa, M. & Zhou, Y . Magnetic skyrmion logic gates: conversion, duplication and merging of skyrmions.Sci. Rep.5, 9400 (2015)

    work page 2015

  21. [21]

    & Otani, Y

    Puebla, J., Kim, J., Kondou, K. & Otani, Y . Spintronic devices for energy-efficient data storage and energy harvesting.Commun. Mater.1, 24 (2020)

    work page 2020

  22. [22]

    M.et al.Skyrmion-based artificial synapses for neuromorphic computing.Nat

    Song, K. M.et al.Skyrmion-based artificial synapses for neuromorphic computing.Nat. Electron.3, 148–155 (2020)

    work page 2020

  23. [23]

    Adv.8, eabq5652 (2022)

    Yokouchi, T.et al.Pattern recognition with neuromorphic computing using magnetic field– induced dynamics of skyrmions.Sci. Adv.8, eabq5652 (2022)

    work page 2022

  24. [24]

    Parkin, S. S. P., Hayashi, M. & Thomas, L. Magnetic domain-wall racetrack memory.Science 320, 190–194 (2008)

    work page 2008

  25. [25]

    & Kubetzka, A

    Spethmann, J., Gr ¨unebohm, M., Wiesendanger, R., von Bergmann, K. & Kubetzka, A. Dis- covery and characterization of a new type of domain wall in a row-wise antiferromagnet.Nat. Commun.12, 3488 (2021)

    work page 2021

  26. [26]

    Ma, C.et al.Electric field-induced creation and directional motion of domain walls and skyrmion bubbles.Nano Lett.19, 353–361 (2018)

    work page 2018

  27. [27]

    Grebenchuk, S.et al.Topological spin textures in an insulating van der Waals ferromagnet. Adv. Mater.36, 2311949 (2024)

    work page 2024

  28. [28]

    & Tretiakov, O

    G ¨obel, B., Mook, A., Henk, J., Mertig, I. & Tretiakov, O. A. Magnetic bimerons as skyrmion analogues in in-plane magnets.Phys. Rev. B99, 060407 (2019). 27

    work page 2019

  29. [29]

    Compact merons and skyrmions in thin chiral magnetic films.Phys

    Ezawa, M. Compact merons and skyrmions in thin chiral magnetic films.Phys. Rev. B83, 100408 (2011)

    work page 2011

  30. [30]

    Capriotti, L., Trumper, A. E. & Sorella, S. Long-range N´eel order in the triangular Heisenberg model.Phys. Rev. Lett.82, 3899 (1999)

    work page 1999

  31. [31]

    White, S. R. & Chernyshev, A. Ne ´el order in square and triangular lattice Heisenberg models. Phys. Rev. Lett.99, 127004 (2007)

    work page 2007

  32. [32]

    Gu, J.et al.Recent advances in convolutional neural networks.Pattern Recognit.77, 354–377 (2018)

    work page 2018

  33. [33]

    Tan, M. & Le, Q. Efficientnet: Rethinking model scaling for convolutional neural networks. InInternational conference on machine learning, 6105–6114 (PMLR, 2019)

    work page 2019

  34. [34]

    Tan, M. & Le, Q. EfficientNetV2: Smaller models and faster training. InInternational con- ference on machine learning, 10096–10106 (PMLR, 2021)

    work page 2021

  35. [35]

    & Sun, J

    He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 770–778 (2016)

    work page 2016

  36. [36]

    MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

    Howard, A. G.et al.MobileNets: Efficient convolutional neural networks for mobile vision applications.arXiv preprint arXiv:1704.04861(2017). 28

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  37. [37]

    & Chen, L.-C

    Sandler, M., Howard, A., Zhu, M., Zhmoginov, A. & Chen, L.-C. MobileNetV2: Inverted residuals and linear bottlenecks. InProceedings of the IEEE conference on computer vision and pattern recognition, 4510–4520 (2018)

    work page 2018

  38. [38]

    & Seyfried, A

    Alia, A., Maree, M., Chraibi, M. & Seyfried, A. A novel V oronoi-based convolutional neural network framework for pushing person detection in crowd videos.Complex Intell. Syst.10, 5005–5031 (2024)

    work page 2024

  39. [39]

    & Chraibi, M

    Alia, A., Maree, M. & Chraibi, M. A hybrid deep learning and visualization framework for pushing behavior detection in pedestrian dynamics.Sensors22, 4040 (2022)

    work page 2022

  40. [40]

    Sci.7, 2000566 (2020)

    Wang, D.et al.Machine learning magnetic parameters from spin configurations.Adv. Sci.7, 2000566 (2020)

    work page 2020

  41. [41]

    G ´omez Albarrac´ın, F. A. & Rosales, H. D. Machine learning techniques to construct detailed phase diagrams for skyrmion systems.Phys. Rev. B105, 214423 (2022)

    work page 2022

  42. [42]

    Y ., Criado, J

    Araz, J. Y ., Criado, J. C. & Spannowsky, M. Identifying magnetic antiskyrmions while they form with convolutional neural networks.J. Magn. Magn. Mater.563, 169806 (2022)

    work page 2022

  43. [44]

    C., Yarımbıyık, A

    Cimen, M., Onel, A. C., Yarımbıyık, A. E., Arikan, M. & Tulum, G. Skyr-net: Deep learning approach for image classification of magnetic structures.J. Magn. Magn. Mater.621, 172912 (2025). 29

    work page 2025

  44. [45]

    & Restrepo-Parra, E

    Salcedo-Gallo, J., Galindo-Gonz ´alez, C. & Restrepo-Parra, E. Deep learning approach for image classification of magnetic phases in chiral magnets.J. Magn. Magn. Mater.501, 166482 (2020)

    work page 2020

  45. [46]

    Jang, Y ., Kim, C. H. & Go, A. Classification of magnetic order from electronic structure by using machine learning.Sci. Rep.13, 12445 (2023)

    work page 2023

  46. [47]

    A.et al.Machine learning magnetism classifiers from atomic coordinates.Iscience 25(2022)

    Merker, H. A.et al.Machine learning magnetism classifiers from atomic coordinates.Iscience 25(2022)

    work page 2022

  47. [48]

    K., Nehete, H., Bindal, N

    Saini, S., Shukla, A. K., Nehete, H., Bindal, N. & Kaushik, B. K. Machine learning-based prediction of antiferromagnetic skyrmion formation.IEEE Trans. Electron Devices71, 2774– 2780 (2024)

    work page 2024

  48. [49]

    I., Swarna, N

    Hasib, F. I., Swarna, N. F. & Alam, M. A. Classification of different magnetic structures from image data using deep neural networks. In2021 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE), 1–6 (IEEE, 2021)

    work page 2021

  49. [50]

    Bokul, S., Shukur, S. S. M. A., Ahmed, S., Bhowmick, T. K. & Alam, M. A. Classification of non-topological magnetic configurations using machine learning. In2020 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE), 1–5 (IEEE, 2020)

    work page 2020

  50. [51]

    & Sinova, J

    Gomonay, O., Jungwirth, T. & Sinova, J. High antiferromagnetic domain wall velocity induced by N´eel spin-orbit torques.Phys. Rev. Lett.117, 017202 (2016)

    work page 2016

  51. [52]

    & Wunderlich, J

    Jungwirth, T., Marti, X., Wadley, P. & Wunderlich, J. Antiferromagnetic spintronics.Nat. Nanotechnol.11, 231–241 (2016). 30

    work page 2016

  52. [53]

    Baltz, V .et al.Antiferromagnetic spintronics.Rev. Mod. Phys.90, 015005 (2018)

    work page 2018

  53. [54]

    Commun.13, 7369 (2022)

    Aldarawsheh, A.et al.Emergence of zero-field non-synthetic single and interchained antifer- romagnetic skyrmions in thin films.Nat. Commun.13, 7369 (2022)

    work page 2022

  54. [55]

    & Lounis, S

    Aldarawsheh, A., Sallermann, M., Abusaa, M. & Lounis, S. A spin model for intrinsic anti- ferromagnetic skyrmions on a triangular lattice.Front. Phys.11, 1175317 (2023)

    work page 2023

  55. [56]

    VFRendering: Python script for visualizing spin textures

    Puerling, T. VFRendering: Python script for visualizing spin textures. https://iffgit.fz-juelich.de/puerling1/VFRendering/-/blob/ 6ec84d9e308d7c97e8dd9563943f11c5e9132aa7/render.py(2023)

    work page 2023

  56. [57]

    & Song, C

    Feng, D., Guan, Z., Wu, X., Wu, Y . & Song, C. Classification of skyrmionic textures and extraction of Hamiltonian parameters via machine learning.Phys. Rev. Appl.21, 034009 (2024)

    work page 2024

  57. [58]

    Younesi, A.et al.A comprehensive survey of convolutions in deep learning: Applications, challenges, and future trends.IEEE Access12, 41180–41218 (2024)

    work page 2024

  58. [59]

    Yamashita, R., Nishio, M., Do, R. K. G. & Togashi, K. Convolutional neural networks: an overview and application in radiology.Insights Imaging9, 611–629 (2018)

    work page 2018

  59. [60]

    & Alemi, A

    Szegedy, C., Ioffe, S., Vanhoucke, V . & Alemi, A. Inception-v4, inception-resnet and the im- pact of residual connections on learning. InProceedings of the AAAI Conference on Artificial Intelligence, vol. 31(1) (2017). 31

    work page 2017

  60. [61]

    & Weinberger, K

    Huang, G., Liu, Z., Van Der Maaten, L. & Weinberger, K. Q. Densely connected convolutional networks. InProceedings of the IEEE conference on computer vision and pattern recognition, 4700–4708 (2017)

    work page 2017

  61. [62]

    InProceedings of the IEEE/CVF international conference on computer vision, 1314–1324 (2019)

    Howard, A.et al.Searching for MobileNetV3. InProceedings of the IEEE/CVF international conference on computer vision, 1314–1324 (2019)

    work page 2019

  62. [63]

    Xception: Deep learning with depthwise separable convolutions

    Chollet, F. Xception: Deep learning with depthwise separable convolutions. InProceedings of the IEEE conference on computer vision and pattern recognition, 1251–1258 (2017)

    work page 2017

  63. [64]

    Ramachandran, P., Zoph, B. & Le, Q. V . Searching for activation functions.arXiv preprint arXiv:1710.05941(2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  64. [65]

    & Sun, G

    Hu, J., Shen, L. & Sun, G. Squeeze-and-excitation networks. InProceedings of the IEEE conference on computer vision and pattern recognition, 7132–7141 (2018)

    work page 2018

  65. [66]

    P.et al.Spirit: Multifunctional framework for atomistic spin simulations.Phys

    M ¨uller, G. P.et al.Spirit: Multifunctional framework for atomistic spin simulations.Phys. Rev. B99, 224414 (2019)

    work page 2019

  66. [67]

    & Tunc, H

    Genc, B. & Tunc, H. Optimal training and test sets design for machine learning.Turk. J. Electr. Eng. Comput. Sci.27, 1534–1545 (2019)

    work page 2019

  67. [68]

    D., Cai, T

    Brown, L. D., Cai, T. T. & DasGupta, A. Interval estimation for a binomial proportion.Sta- tistical science16, 101–133 (2001). 32

    work page 2001

  68. [69]

    R.et al.Grad-cam: Visual explanations from deep networks via gradient-based localization

    Selvaraju, R. R.et al.Grad-cam: Visual explanations from deep networks via gradient-based localization. InProceedings of the IEEE international conference on computer vision, 618– 626 (2017)

    work page 2017

  69. [70]

    Vansteenkiste, A.et al.The design and verification of MuMax3.AIP Adv.4(2014). 33

    work page 2014