pith. sign in

arxiv: 2512.04024 · v1 · submitted 2025-12-03 · ⚛️ physics.comp-ph · cond-mat.stat-mech· cond-mat.supr-con

Predicting parameters of a model cuprate superconductor using machine learning

V. A. Ulitko , D. N. Yasinskaya , S. A. Bezzubin , A. A. Koshelev , Y. D. Panov This is my paper

Pith reviewed 2026-05-17 01:47 UTC · model grok-4.3

classification ⚛️ physics.comp-ph cond-mat.stat-mechcond-mat.supr-con
keywords parameterslearningmodelphasediagramsmachinecorrespondcuprate
0
0 comments X

The pith

An adapted U-Net model trained on mean-field phase diagrams accurately predicts Hamiltonian parameters for a cuprate superconductor when validated on Monte Carlo simulation data.

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

In condensed matter physics, cuprate superconductors are modeled with Hamiltonians that have several adjustable parameters describing electron behavior and interactions. Computing the full phase diagram for many different parameter combinations is extremely time-consuming, limiting how well models can be matched to real experiments. This work turns the problem around by using machine learning to guess the parameters from a given phase diagram. The authors created a large set of phase diagrams using a faster but approximate mean-field method to train neural networks. They tested three architectures and found that a modified U-Net, normally used for image tasks, worked best when repurposed to output numerical parameter values. After training, they checked the model on phase diagrams generated by a more accurate but slower Monte Carlo simulation technique. The network recovered the original parameters well in most cases. Importantly, it performed worse only in regions where changing the parameters produced little visible change in the phase diagram, showing that the model was responding to actual physical sensitivities rather than artifacts. This approach offers a way to extract which parameters matter most and to speed up analysis of complex models.

Core claim

It is shown that the model accurately predicts all considered Hamiltonian parameters, and areas of low prediction accuracy correspond to regions of parametric insensitivity in the phase diagrams.

Load-bearing premise

The phase diagrams generated within the mean-field approximation contain sufficient structural information for a neural network trained on them to generalize accurately to phase diagrams produced by semi-classical heat bath Monte Carlo simulations.

Figures

Figures reproduced from arXiv: 2512.04024 by A. A. Koshelev, D. N. Yasinskaya, S. A. Bezzubin, V. A. Ulitko, Y. D. Panov.

Figure 1
Figure 1. Figure 1: Comparison of the results from MFA and numerical modeling using the heat [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Two-stage neural network operation diagram. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The distribution histograms of the parameters [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Bar charts of RMSE and R2 metrics for each parameter ∆, V , tb, tp for MFA [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: This effect does not necessarily indicate mistakes in the machine learn￾ing model. Several factors could account for this. First, there might be an imbalance in the training dataset. Although the scatter plots display hori￾11 [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 6
Figure 6. Figure 6: Scatter plots show correlation between predicted and expected values for each [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Sensitivity analysis of the T–n phase diagrams to parameter variations: (a) a series of diagrams plotted for fixed ∆, V , and varying tb ∈ [0.02 − 0.22]; (b) a series of diagrams plotted for fixed ∆, tb, and varying V ∈ [0.02 − 0.18]. Different colors represent different homogeneous phases. Light grey indicates the NO phase, red represents AFM, green represents CO, blue represents BS, and dark grey represe… view at source ↗
Figure 6
Figure 6. Figure 6: This suggests that the machine learning model acts not merely as a “black box”, but as a tool capable of analyzing the physical properties of the un￾derlying physical model. The neural network’s inability to predict certain parameters indicates their weak influence on the target variable (the phase diagram) within the considered range. Conversely, it successfully identifies parameter intervals where the pa… view at source ↗
Figure 8
Figure 8. Figure 8: Scatter plots of the predicted and expected parameter values for the heat bath [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
read the original abstract

The computational complexity of calculating phase diagrams for multi-parameter models significantly limits the ability to select parameters that correspond to experimental data. This work presents a machine learning method for solving the inverse problem - forecasting the parameters of a model Hamiltonian for a cuprate superconductor based on its phase diagram. A comparative study of three deep learning architectures was conducted: VGG, ResNet, and U-Net. The latter was adapted for regression tasks and demonstrated the best performance. Training the U-Net model was performed on an extensive dataset of phase diagrams calculated within the mean-field approximation, followed by validation on data obtained using a semi-classical heat bath algorithm for Monte Carlo simulations. It is shown that the model accurately predicts all considered Hamiltonian parameters, and areas of low prediction accuracy correspond to regions of parametric insensitivity in the phase diagrams. This allows for the extraction of physically interpretable patterns and validation of the significance of parameters for the system. The results confirm the promising potential of applying machine learning to analyze complex physical models in condensed matter physics.

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

3 major / 2 minor

Summary. The manuscript presents a machine-learning approach to the inverse problem of recovering Hamiltonian parameters for a model cuprate superconductor from its phase diagram. A U-Net architecture adapted for regression is trained on an extensive set of phase diagrams generated in the mean-field approximation and then validated on diagrams produced by semi-classical heat-bath Monte Carlo simulations. The central claim is that the trained model accurately recovers all considered parameters and that regions of lower prediction accuracy coincide with physically insensitive regions of the phase diagram.

Significance. If the reported generalization from mean-field training data to Monte Carlo validation data holds with quantitative fidelity, the method would offer a practical route to rapid parameter estimation for multi-parameter models whose direct fitting is computationally prohibitive. The explicit linkage of prediction error to parametric insensitivity is a positive feature that could aid physical interpretation. However, the absence of side-by-side quantitative metrics comparing mean-field and Monte Carlo diagrams leaves the transferability assumption untested, limiting the immediate impact.

major comments (3)
  1. [Abstract / Results] The abstract and strongest claim assert that the U-Net 'accurately predicts all considered Hamiltonian parameters' on Monte Carlo validation data, yet no numerical performance measures (RMSE, MAE, or correlation coefficients) or dataset sizes are supplied. Without these, it is impossible to judge whether the accuracy is sufficient to support the inverse-problem application.
  2. [Methods / Validation procedure] The central generalization step—training exclusively on mean-field phase diagrams and validating on semi-classical Monte Carlo diagrams—rests on an unverified assumption of structural similarity. Mean-field theory suppresses fluctuations that shift transition lines and alter the topology of ordered regions; no pixel-wise discrepancy, order-parameter profile comparison, or feature-alignment metric between the two classes of diagrams is reported. This directly affects the load-bearing claim that the network recovers physically meaningful parameters.
  3. [Results / Discussion] The statement that low-accuracy regions 'correspond to regions of parametric insensitivity' is presented as a physical insight, but the manuscript does not show an explicit sensitivity analysis (e.g., finite-difference derivatives of the phase diagram with respect to each Hamiltonian parameter) that would independently confirm this correspondence.
minor comments (2)
  1. [Model Hamiltonian] Notation for the Hamiltonian parameters and the precise definition of the order parameters used to construct the phase diagrams should be stated explicitly in the main text rather than deferred to supplementary material.
  2. [Figures] Figure captions should include the number of training/validation samples and the precise Monte Carlo parameters (lattice size, number of sweeps, etc.) so that reproducibility is immediate.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. These have helped us identify areas where additional quantitative support and clarification will strengthen the manuscript. We address each major comment below and will make the indicated revisions.

read point-by-point responses

  1. Referee: [Abstract / Results] The abstract and strongest claim assert that the U-Net 'accurately predicts all considered Hamiltonian parameters' on Monte Carlo validation data, yet no numerical performance measures (RMSE, MAE, or correlation coefficients) or dataset sizes are supplied. Without these, it is impossible to judge whether the accuracy is sufficient to support the inverse-problem application.

    Authors: We agree that explicit numerical metrics are necessary to allow readers to evaluate the claimed accuracy. In the revised manuscript we will report RMSE, MAE, and Pearson correlation coefficients for each predicted Hamiltonian parameter on the Monte Carlo validation set. We will also state the sizes of the mean-field training set and the Monte Carlo validation set. These values will be added to the Results section and referenced concisely in the abstract. revision: yes

  2. Referee: [Methods / Validation procedure] The central generalization step—training exclusively on mean-field phase diagrams and validating on semi-classical Monte Carlo diagrams—rests on an unverified assumption of structural similarity. Mean-field theory suppresses fluctuations that shift transition lines and alter the topology of ordered regions; no pixel-wise discrepancy, order-parameter profile comparison, or feature-alignment metric between the two classes of diagrams is reported. This directly affects the load-bearing claim that the network recovers physically meaningful parameters.

    Authors: We acknowledge that a direct quantitative comparison would make the transferability assumption more transparent. The fact that the model recovers parameters accurately on Monte Carlo data already supplies indirect support for sufficient structural overlap, yet we agree this is not a substitute for explicit metrics. In the revision we will add a short subsection presenting pixel-wise discrepancy maps and order-parameter profile comparisons between representative mean-field and Monte Carlo diagrams, thereby documenting the degree of similarity that enables successful generalization. revision: yes

  3. Referee: [Results / Discussion] The statement that low-accuracy regions 'correspond to regions of parametric insensitivity' is presented as a physical insight, but the manuscript does not show an explicit sensitivity analysis (e.g., finite-difference derivatives of the phase diagram with respect to each Hamiltonian parameter) that would independently confirm this correspondence.

    Authors: We accept that an explicit sensitivity analysis would provide independent confirmation of the claimed correspondence. In the revised manuscript we will include a finite-difference sensitivity study: for each Hamiltonian parameter we will compute the change in selected phase-diagram features (transition temperatures, ordered-region areas) under small parameter perturbations and overlay the resulting sensitivity maps with the observed prediction-error maps. This analysis will be presented in the Discussion to substantiate the physical interpretation. revision: yes

Circularity Check

0 steps flagged

No circularity: ML inversion trained on mean-field diagrams and validated on independent Monte Carlo diagrams

full rationale

The paper generates phase diagrams under the mean-field approximation to train the U-Net regressor and then tests recovery of the same Hamiltonian parameters on diagrams produced by a separate semi-classical heat-bath Monte Carlo algorithm. This constitutes an external benchmark rather than a self-referential fit. No equation or claim reduces a predicted parameter to a quantity defined inside the training set, no self-citation supplies a load-bearing uniqueness theorem, and no ansatz is smuggled through prior work. The reported accuracy and the correspondence between low-accuracy regions and parametric insensitivity are therefore independent of the training data by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the mean-field approximation generating representative training data and on the assumption that phase diagram features encode sufficient information for parameter inference; no free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption Phase diagrams computed in the mean-field approximation contain sufficient information to train a model that predicts parameters accurately for more precise simulations.
    Training is performed exclusively on mean-field data as stated in the abstract.
  • domain assumption The relationship between Hamiltonian parameters and phase diagram features is learnable by the neural network architecture.
    The reported success of the U-Net regression implies this mapping exists and is captured.

pith-pipeline@v0.9.0 · 5509 in / 1470 out tokens · 62119 ms · 2026-05-17T01:47:56.800156+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages · 1 internal anchor

  1. [1]

    A. S. Moskvin, True charge-transfer gap in parent insu- lating cuprates, Physical Review B 84 (7) (2011) 075116. doi:10.1103/PhysRevB.84.075116

    work page doi:10.1103/physrevb.84.075116 2011

  2. [2]

    doi:10.1088/0953-8984/25/8/085601

    A.Moskvin, Perspectivesofdisproportionationdrivensuperconductivity in strongly correlated 3d compounds, Journal of Physics: Condensed Matter 25 (8) (2013) 085601. doi:10.1088/0953-8984/25/8/085601

    work page doi:10.1088/0953-8984/25/8/085601 2013

  3. [3]

    Moskvin, Y

    A. Moskvin, Y. Panov, Model of charge triplets for high-Tc cuprates, Journal of Magnetism and Magnetic Materials 550 (2022) 169004. doi:10.1016/j.jmmm.2021.169004

    work page doi:10.1016/j.jmmm.2021.169004 2022

  4. [4]

    Y. D. Panov, Critical Temperatures of a Model Cuprate, Physics of Metals and Metallography 120 (13) (2019) 1276–1281. doi:10.1134/S0031918X19130222

    work page doi:10.1134/s0031918x19130222 2019

  5. [5]

    A. S. Moskvin, Y. D. Panov, Nature of the Pseudogap Phase of HTSC Cuprates, Physics of the Solid State 62 (9) (2020) 1554–1561. doi:10.1134/S1063783420090206

    work page doi:10.1134/s1063783420090206 2020

  6. [6]

    Moskvin, Y

    A. Moskvin, Y. Panov, Effective-Field Theory for Model High-Tc Cuprates, Condensed Matter 6 (3) (2021) 24. doi:10.3390/condmat6030024. 18

    work page doi:10.3390/condmat6030024 2021

  7. [7]

    Y. D. Panov, A. S. Moskvin, A. A. Chikov, V. A. Ulitko, Monte carlo simulation of a model cuprate, Journal of Physics: Conference Series 2043 (1) (2021) 012007. doi:10.1088/1742-6596/2043/1/012007

    work page doi:10.1088/1742-6596/2043/1/012007 2043

  8. [8]

    Ulitko, Y

    V. Ulitko, Y. D. Panov, A. Moskvin, Classical monte carlo algorithm for simulation of a pseudospin model for cuprates, arXiv preprint arXiv:2301.11708 (2023). doi:10.48550/arXiv.2301.11708

    work page doi:10.48550/arxiv.2301.11708 2023

  9. [9]

    J. Lund, H. Wang, R. Braatz, R. García, Machine learning of phase diagrams, Materials Advances 3 (23) (2022) 8485–8497. doi:10.1039/D2MA00524G

    work page doi:10.1039/d2ma00524g 2022

  10. [10]

    Casert, T

    C. Casert, T. Vieijra, J. Nys, J. Ryckebusch, Interpretable ma- chine learning for inferring the phase boundaries in a nonequi- librium system, Physical Review E 99 (2) (2019) 023304. doi:10.1103/PhysRevE.99.023304

    work page doi:10.1103/physreve.99.023304 2019

  11. [11]

    & Laino, T

    F. Zipoli, V. Viterbo, O. Schilter, L. Kahle, T. Laino, Prediction of phase diagrams and associated phase structural properties, Indus- trial & Engineering Chemistry Research 61 (24) (2022) 8378–8389. doi:10.1021/acs.iecr.2c00355

    work page doi:10.1021/acs.iecr.2c00355 2022

  12. [12]

    D. Bayo, B. Çivitcioğlu, J. J. Webb, A. Honecker, R. A. Römer, Machine learning of phases and structures for model systems in physics, Journal of the Physical Society of Japan 94 (3) (2025) 031002. doi:10.7566/JPSJ.94.031002

    work page doi:10.7566/jpsj.94.031002 2025

  13. [13]

    A., Ahmadi, E

    M. Aghaaminiha, S. A. Ghanadian, E. Ahmadi, A. M. Farnoud, A machine learning approach to estimation of phase dia- grams for three-component lipid mixtures, Biochimica et Bio- physica Acta (BBA)-Biomembranes 1862 (9) (2020) 183350. doi:10.1016/j.bbamem.2020.183350

    work page doi:10.1016/j.bbamem.2020.183350 2020

  14. [14]

    J. Sun, J. Qu, C. Zhao, X. Zhang, X. Liu, J. Wang, C. Wei, X. Liu, M. Wang, P. Zeng, et al., Precise prediction of phase-separation key residues by machine learning, Nature Communications 15 (1) (2024)

    work page 2024

  15. [15]

    doi:10.1038/s41467-024-46901-9

    work page doi:10.1038/s41467-024-46901-9

  16. [16]

    E. P. Van Nieuwenburg, Y.-H. Liu, S. D. Huber, Learning phase transitions by confusion, Nature Physics 13 (5) (2017) 435–439. doi:10.1038/nphys4037. 19

    work page doi:10.1038/nphys4037 2017

  17. [17]

    Zhang, J

    W. Zhang, J. Liu, T.-C. Wei, Machine learning of phase transitions in the percolation and xy models, Physical Review E 99 (3) (2019) 032142. doi:10.1103/PhysRevE.99.032142

    work page doi:10.1103/physreve.99.032142 2019

  18. [18]

    Canabarro, F

    A. Canabarro, F. F. Fanchini, A. L. Malvezzi, R. Pereira, R. Chaves, Unveiling phase transitions with machine learning, Physical Review B 100 (4) (2019) 045129. doi:10.1103/PhysRevB.100.045129

    work page doi:10.1103/physrevb.100.045129 2019

  19. [19]

    H. Y. Kwon, H. Yoon, C. Lee, G. Chen, K. Liu, A. Schmid, Y. Wu, J. Choi, C. Won, Magnetic hamiltonian parameter estimation using deep learning techniques, Science advances 6 (39) (2020) eabb0872. doi:10.1126/sciadv.abb0872

    work page doi:10.1126/sciadv.abb0872 2020

  20. [20]

    F. A. Gómez Albarracín, H. D. Rosales, Machine learning techniques to construct detailed phase diagrams for skyrmion systems, Physical Review B 105 (21) (2022) 214423. doi:10.1103/PhysRevB.105.214423

    work page doi:10.1103/physrevb.105.214423 2022

  21. [21]

    Y. Che, C. Gneiting, T. Liu, F. Nori, Topological quantum phase transi- tions retrieved through unsupervised machine learning, Physical Review B 102 (13) (2020) 134213. doi:10.1103/PhysRevB.102.134213

    work page doi:10.1103/physrevb.102.134213 2020

  22. [22]

    Stanev, C

    V. Stanev, C. Oses, A. G. Kusne, E. Rodriguez, J. Paglione, S. Cur- tarolo, I. Takeuchi, Machine learning modeling of superconducting crit- ical temperature, npj Computational Materials 4 (1) (2018) 29

    work page 2018

  23. [23]

    Zhang, X

    Y. Zhang, X. Xu, Predicting the superconducting transition tempera- ture of high-temperature layered superconductors via machine learning, Physica C: Superconductivity and its Applications 595 (2022) 1354031. doi:10.1016/j.physc.2022.1354031

    work page doi:10.1016/j.physc.2022.1354031 2022

  24. [24]

    S. Xie, Y. Quan, A. Hire, B. Deng, J. DeStefano, I. Salinas, U. Shah, L. Fanfarillo, J. Lim, J. Kim, et al., Machine learning of superconducting critical temperature from eliashberg theory, npj Computational Materi- als 8 (1) (2022) 14. doi:10.1038/s41524-021-00666-7

    work page doi:10.1038/s41524-021-00666-7 2022

  25. [25]

    P. J. G. Nieto, E. G. Gonzalo, L. A. M. García, L. Álvarez-de Prado, A. B. Sánchez, Predicting the critical superconducting temperature us- ing the random forest, mlp neural network, m5 model tree and multivari- ate linear regression, Alexandria Engineering Journal 86 (2024) 144–156. doi:10.1016/j.aej.2023.11.034. 20

    work page doi:10.1016/j.aej.2023.11.034 2024

  26. [26]

    Y. D. Panov, A. S. Moskvin, A. A. Chikov, I. L. Avvakumov, Compe- tition of Spin and Charge Orders in a Model Cuprate, J. Supercond. Novel Magn. 29 (4) (2016) 1077–1083. doi:10.1007/s10948-016-3378-5

    work page doi:10.1007/s10948-016-3378-5 2016

  27. [27]

    Micnas, J

    R. Micnas, J. Ranninger, S. Robaszkiewicz, Superconductiv- ity in narrow-band systems with local nonretarded attractive interactions, Reviews of Modern Physics 62 (1) (1990) 113. doi:10.1103/RevModPhys.62.113

    work page doi:10.1103/revmodphys.62.113 1990

  28. [28]

    Fradkin, S

    E. Fradkin, S. A. Kivelson, J. M. Tranquada, Colloquium : Theory of in- tertwined orders in high temperature superconductors, Reviews of Mod- ern Physics 87 (2) (2015) 457–482. doi:10.1103/RevModPhys.87.457

    work page doi:10.1103/revmodphys.87.457 2015

  29. [29]

    Miyatake, M

    Y. Miyatake, M. Yamamoto, J. Kim, M. Toyonaga, O. Nagai, On the implementation of the’heat bath’algorithms for monte carlo simulations of classical heisenberg spin systems, Journal of Physics C: solid state physics 19 (14) (1986) 2539. doi:10.1088/0022-3719/19/14/020

    work page doi:10.1088/0022-3719/19/14/020 1986

  30. [30]

    Y. D. Panov, S. V. Nuzhin, V. S. Ryumshin, A. S. Moskvin, The monte carlo method for the orthonickelate model, Fizika Tverdogo Tela 66 (7) (2024) 1221–1227

    work page 2024

  31. [31]

    Y. D. Panov, V. A. Ulitko, D. N. Yasinskaya, A. S. Moskvin, Mod- ified monte carlo method with the heat bath algorithm for a model cuprate, arXiv preprint arXiv:2510.08289v1 (2025). doi:10.1088/1742- 6596/2043/1/012007

    work page doi:10.1088/1742- 2025

  32. [32]

    Very Deep Convolutional Networks for Large-Scale Image Recognition

    K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, arXiv preprint arXiv:1409.1556 (2014). doi:10.48550/arXiv.1409.1556

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1409.1556 2014

  33. [33]

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

    work page 2016

  34. [34]

    U-Net: Convolutional networks for biomedical image segmentation

    O. Ronneberger, P. Fischer, T. Brox, U-net: Convolutional networks for biomedicalimagesegmentation, in: InternationalConferenceonMedical image computing and computer-assisted intervention, Springer, 2015, pp. 234–241. doi:10.1007/978-3-319-24574-4_28. 21

    work page doi:10.1007/978-3-319-24574-4_28 2015

  35. [35]

    R. Azad, E. K. Aghdam, A. Rauland, Y. Jia, A. H. Avval, A. Bo- zorgpour, S. Karimijafarbigloo, J. P. Cohen, E. Adeli, D. Merhof, Medical image segmentation review: The success of u-net, IEEE Transactions on Pattern Analysis and Machine Intelligence (2024). doi:10.1109/TPAMI.2024.3435571

    work page doi:10.1109/tpami.2024.3435571 2024

  36. [36]

    Lee, W.-T

    W.-J. Lee, W.-T. Hsieh, B.-H. Fang, K.-H. Kao, N.-Y. Chen, De- vice simulations with a u-net model predicting physical quantities in two-dimensional landscapes, Scientific reports 13 (1) (2023) 731. doi:10.1038/s41598-023-27599-z

    work page doi:10.1038/s41598-023-27599-z 2023

  37. [37]

    J. Ye, Y. Huang, K. Liu, U-net based vortex detection in bose–einstein condensates with automatic correction for manually mislabeled data, Scientific Reports 13 (1) (2023) 21278. doi:10.1038/s41598-023-48719-9

    work page doi:10.1038/s41598-023-48719-9 2023

  38. [38]

    Kamali, K

    A. Kamali, K. Laksari, Physics-informed unets for discovering hidden elasticity in heterogeneous materials, Journal of the me- chanical behavior of biomedical materials 150 (2024) 106228. doi:10.1016/j.jmbbm.2023.106228

    work page doi:10.1016/j.jmbbm.2023.106228 2024

  39. [39]

    Chen, Y.-S

    B.-R. Chen, Y.-S. Hsiao, W.-C. Lin, W.-J. Lee, N.-Y. Chen, T.-L. Wu, Using u-net convolutional neural network to model pixel-based electro- static potential distributions in gan power mis-hemts, Scientific reports 14 (1) (2024) 8151. doi:10.1038/s41598-024-58112-9

    work page doi:10.1038/s41598-024-58112-9 2024

  40. [40]

    W. Yao, Z. Zeng, C. Lian, H. Tang, Pixel-wise regression using u-net and its application on pansharpening, Neurocomputing 312 (2018) 364–371

    work page 2018

  41. [41]

    Gertsvolf, M

    D. Gertsvolf, M. Horvat, D. Aslam, A. Khademi, U. Berardi, A u-net convolutional neural network deep learning model application for identi- fication of energy loss in infrared thermographic images, Applied Energy 360 (2024) 122696. 22

    work page 2024