Pith. sign in

REVIEW 3 major objections 5 minor 80 references

Transition of AI Models in dependence of noise

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The authors show that the accuracy-versus-noise transition narrows as a power law in model size, with exponent 0.27 over three decades.

desk verdict A potentially useful empirical scaling law, undermined by a bad noise conversion and a lack of reproducibility. read the letter →

arxiv 2506.16715 v3 pith:TV6EVQNS submitted 2025-06-20 physics.data-an

Thomas Seidler , Markus Abel This is my paper · ORCID
classification physics.data-an
keywords cognitiontransitionfinite-sizescalinglawscriticalexponentsuniversalityclassEfficientNetimageclassificationGaussiannoise
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that the loss of image-classification accuracy under added Gaussian noise is a sharp "cognition transition" and that, as models grow, this transition narrows according to a finite-size scaling law. Using the EfficientNet family (roughly $10^4$ to $10^7$ parameters), the authors corrupt test images with a diffusion-style forward noise process and record accuracy as a function of noise step. They find a power-law narrowing of the transition width with a slope of about 0.27 across three decades of model size, with a common critical noise for all sizes. If correct, this gives a practical recipe: measure small models cheaply and extrapolate to larger ones, and it suggests that statistical-mechanics universality classes could organize model robustness.

What carries the argument

The central object is the finite-size scaling ansatz $F(L) \propto L^{-\lambda} f((T-T_c)L^{1/\nu})$, applied with $L$ the number of trainable parameters and $T$ the diffusion noise step. The transition width is extracted with the level-set method: nested accuracy thresholds such as $(0.1,0.9)$ and $(0.2,0.8)$ define noise-step intervals, and the interval width $\Delta = |\beta(a_{\text{up}})-\beta(a_{\text{low}})|$ is plotted against model size. Noise is injected through the forward process of denoising diffusion probabilistic models with the linear schedule $\beta_t = 10^{-4} + 4\times 10^{-6} t$ over 5000 steps, giving Gaussian corruption with analytic marginals. This machinery produces the central plot: a double-logarithmic straight line of slope 0.27 for transition width versus parameter count.

What would settle it

Apply the same noise schedule and level-set width measurement to a second architecture family; if the log-log plot of transition width versus parameter count is not a straight line with a stable slope, the claimed universal scaling fails. As an arithmetic check, compute $\sigma = \sqrt{1-\bar\alpha_t}$ from the stated schedule at $t_c = 1845.934$ and compare $\log \sigma$ with the quoted value 11.

Watch

Extended reading notes

Core claim

The central discovery is that the transition from good classification to chance-level performance under controlled Gaussian noise behaves like a finite-size critical phenomenon. For a fixed family of models, larger networks perform better and the accuracy transition becomes sharper, while the critical noise level remains fixed (the authors report $t_c = 1845.934$ in diffusion steps for all EfficientNet sizes). The sharpness, defined as the width $\Delta$ of nested accuracy level sets, is claimed to obey a power law in the number of parameters, $\Delta \sim N^{-0.27}$, across three decades. The authors interpret this as evidence for the finite-size scaling form $F(L) \sim L^{-\lambda} f((T-T_c)L^{1/\nu})$ and take it to indicate universal behavior within a class of similarly built models.

Load-bearing premise

The argument assumes that accuracy-versus-noise curves for models of different sizes all line up according to one statistical scaling law, with a particular step-counting scheme for Gaussian noise as the measure of noise; this has been tested only on one model family.

Editorial extensions

If this is right

  • If the power law is correct, a practitioner can measure the transition on a small, cheaply trained model and extrapolate to the model size needed for a target robustness, avoiding full large-scale training.
  • The finding that the critical noise is the same across EfficientNet sizes implies that scaling up model size sharpens the transition rather than moving where failure begins.
  • The reported exponent 0.27 gives a quantitative baseline: model families in the same universality class should reproduce this slope, and deviations would signal a different class.
  • The level-set width protocol offers a systematic way to compare robustness across models without depending on the exact accuracy at any one noise level.

Reading between the lines

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

  • If the scaling extends beyond EfficientNet, applying the same protocol to other architecture families would be a direct test; a shared exponent would mean robustness to Gaussian noise is governed mainly by parameter count, not by architectural detail.
  • The noise axis itself deserves scrutiny: the quoted $\log \sigma_c \approx 11$ does not match the analytic $\sigma = \sqrt{1-\bar\alpha_t}$ from Eq. 5 at the reported $t_c = 1845.934$, so the quantitative position of the transition may need a corrected conversion.
  • Beyond classification accuracy, the same finite-size scaling could be probed on regression tasks or calibration error to ask whether the cognition transition and its exponent are universal across tasks.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper studies the accuracy of image classifiers (EfficientNet family, both ImageNet-pretrained B0–B7 and custom small models trained on CIFAR100) as a function of Gaussian noise added via the DDPM forward process. The authors identify a sharp “cognition transition” from high accuracy to near-zero accuracy when the noise step t reaches a critical value, and they report that the width of this transition narrows with increasing model size according to a power law with slope ≈ 0.27 over approximately three decades of model size. They interpret this as finite-size scaling, claim a unique critical noise for all models, and suggest that the result can be used to extrapolate robustness to larger models without full training. The paper is written as an empirical physics-style scaling study.

Significance. If the reported power law were correct, it would provide a quantitative finite-size scaling relation for classifier robustness under additive Gaussian noise, with practical value for model sizing and an interesting bridge between statistical mechanics and deep learning. The paper also attempts to use a level-set method for measuring transition widths, which, if validated, could be a useful tool. However, the central quantitative claim rests on the choice of the DDPM step index t as the noise variable, and the manuscript contains an internal inconsistency in the conversion between t and the actual noise standard deviation sigma. Because the scaling may be an artifact of the t-to-sigma saturation, the significance of the result, while potentially high, is not currently established. Strengths include the use of a controlled, well-defined noise process (DDPM) and a clear statement of the finite-size scaling hypothesis, but the analysis lacks error bars, seed variability, and a data-collapse test.

major comments (3)
  1. [Section IV, Fig. 2 and Eqs. (2)–(5)] The claim that the unique critical step t_c = 1845.934 corresponds to log(sigma_c) ≈ 11 is internally inconsistent with the stated noise schedule. Using beta_t = 1e-4 + 4e-6 t, the cumulative variance at t_c is sum_{s=0}^{t_c} beta_s ≈ 7.0, giving alpha_bar(t_c) ≈ e^{-7} and sigma(t_c) = sqrt(1 - alpha_bar) ≈ 0.9991, so log(sigma_c) is near 0, not 11. More importantly, sigma(t) is a saturating function of t: between t = 1000 and t = 2000 it changes from about 0.94 to 0.9998. Thus the transition region in t spans only a tiny interval of actual noise amplitude sigma, and a power law in the step index t does not imply a power law in the physical noise level. The claimed “three decades of noise” (lead paragraph) is actually three decades of model size, and the quantitative core of the paper—the width scaling—is measured in the proxy variable t without demonstrating that it survives conversion to sigma.
  2. [Section II and Fig. 4] The finite-size scaling form F(L) ∝ L^{-lambda} f((T - T_c) L^{1/nu}) is asserted but never tested by a data collapse. The exponent 0.27 is obtained from a log-log fit of transition width versus parameter count, but no error bars, number of seeds, or uncertainty analysis are reported, and the fit range includes only about ten model sizes. A genuine scaling hypothesis should collapse all the accuracy-versus-noise curves onto a single master curve when plotted against (t - t_c) L^{1/nu}; this is not shown. Without such a collapse, the reported power law could reflect the specific choice of level sets or the accidental alignment of a small number of points, rather than a universal scaling relation.
  3. [Section IV (Fig. 2 inset)] The assertion that there is a single critical noise value t_c ≈ 1845.934 for all EfficientNet models is stated without any description of how this value was determined. In standard finite-size scaling, the effective transition point typically shifts with system size, approaching the thermodynamic limit as L grows. If the transition center is claimed to be exactly size-independent, that is a special and non-generic property that needs justification; otherwise the nested-interval level-set measurement of the width may be dominated by the arbitrarily chosen accuracy levels rather than by a well-defined scaling variable. The uniqueness of t_c is load-bearing for the interpretation of the width scaling, and it is asserted rather than demonstrated.
minor comments (5)
  1. [Lead paragraph and Section IV, Fig. 3] The lead paragraph states that the authors “can determine scaling behavior over three decades of noise,” but the scaling in Fig. 4 is over model size (parameter count). Please clarify which quantity spans the three decades; if both are intended, the text should distinguish them explicitly.
  2. [Section III.B.1, Eq. (6)] In the accuracy definition, the text defines y as the predicted label and ŷ as the true label, but the indicator function is written as 1(ŷ_i = y_i), which is symmetric. Please correct the notation so that the roles of predicted and true labels are unambiguous.
  3. [Section III.B] The description of the custom small models as “EfficientNet-compatible sizes” is vague. Please specify the exact architecture (depth, width, resolution) and training hyperparameters for these models, since they form the majority of the data points in the scaling fit.
  4. [Table I] The FLOPS column appears to use units of GFLOPs (or similar), but the units are not stated. Please add explicit units and a brief definition of how FLOPS is computed.
  5. [General] Several references in the bibliography (e.g., references to MNIST and Fashion-MNIST in the reference list) are not cited in the text. Please ensure that all listed references are actually used and cited where relevant.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the power-law exponent is an empirical fit whose inputs do not include the claimed result, and the only self-citation is not load-bearing.

full rationale

The paper's central claim—a power-law narrowing of the cognition-transition width with model size—is obtained by first defining transition widths Δ from measured accuracy curves via level sets and then fitting log Δ versus log N in Fig. 4. Nothing in the definition of Δ or in the level-set procedure presupposes the power law or fixes its exponent; the exponent 0.27 is a regression output, not a pre-imposed input. The proposed model-size estimation is an extrapolation of that fitted law to larger sizes, not a prediction of a quantity already used as a fit target, so it does not reduce to the fit. The finite-size scaling form F(L) ∝ L^{-λ} f((T−Tc)L^{1/ν}) is borrowed from standard statistical mechanics and used as an organizing assumption; using an assumed scaling form to analyze data is not circular, even if untested. The self-citation to the authors' earlier ICAART paper supports only the generic expectation that larger models show sharper transitions, an expectation that is independently standard in finite-size scaling theory, and the measurements do not rest on that citation. The internal inconsistencies noted by the skeptical reading—t_c≈1845.934 versus 'log(σ_c)≈11' and the axis label 'log(step)∈(10,12)'—are correctness issues about the mapping from diffusion step to noise amplitude, not circularity: they do not make any derived quantity equal to an input by construction. The paper also explicitly acknowledges its limitation that it cannot yet identify which other models share the universality class, further indicating that the scaling claim is presented as an empirical finding rather than as a consequence of its own definitions.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim (power-law scaling of transition width) depends on a fitted exponent, a hand-chosen noise schedule, and the untested analogy to statistical-mechanical finite-size scaling.

free parameters (3)
  • transition width scaling exponent alpha = 0.27
    Slope of log(Delta) vs log(N) fit over three decades (Fig. 4); reported as the central scaling result but not derived from theory.
  • DDPM noise schedule constants = beta_0 = 1e-4, slope = 4e-6, T = 5000
    Hand-chosen linear schedule rescaled from Ho et al. (2020). The critical step t_c and transition widths are measured in these units, so the exponent depends on this arbitrary choice.
  • accuracy level sets = e.g., (0.1, 0.9), (0.2, 0.8)
    Chosen thresholds for defining transition width; the paper checks consistency across levels but the choice is arbitrary.
assumptions (4)
  • domain assumption Accuracy of a classifier behaves as an order parameter analogous to magnetization, with noise playing the role of temperature.
    Section II (Background) draws this analogy; the entire finite-size scaling analysis rests on it.
  • ad hoc to paper The transition width follows the finite-size scaling form F(L) ~ L^{-lambda} f((T-Tc)L^{1/nu}).
    Stated in Section II and used implicitly to fit the power law in Fig. 4; no derivation or independent test.
  • domain assumption The DDPM forward process with linear schedule produces noise levels that are a meaningful monotonic proxy for data corruption.
    Section III.A adopts the DDPM perturbation scheme; the reported log(sigma_c) is inconsistent with Eq. 5, so this conversion is suspect.
  • ad hoc to paper EfficientNet variants of different sizes, plus the custom small models, belong to a single universality class.
    Section V admits other models are untested; the power-law fit assumes a common scaling function.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transition of AI Models in dependence of noise." pith.science (2026). https://pith.science/paper/TV6EVQNS

@misc{pith2026250616715,
  author       = {Pith},
  title        = {Pith review of: Transition of AI Models in dependence of noise},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TV6EVQNS}},
  note         = {Machine review of arXiv:2506.16715}
}
read the original abstract

We investigate the dependence of the score on noise in the data, and on the network size. As a result, we obtain the so-called "cognition transition" from good performance to zero with increasing noise. The understanding of this transition is of fundamental scientific and practical interest. We use concepts from statistical mechanics to understand how a changing finite size of models affects the cognition ability under the presence or corrupted data. On one hand, we study if there is a universal aspect in the transition to several models, on the other hand we go into detail how the approach of the cognition transition point can be captured quantitatively. Therefore, we use the so-called scaling approach from statistical mechanics and find a power-law behaviour of the transition width with increasing model size. Since our study is aimed at universal aspects we use well-know models and data for image classification. That way we avoid uncertainties in data handling or model setup. The practical implication of our results is a tool to estimate model sizes for a certain "universality class" of models, without the need to investigate large sizes, just by extrapolating the scaling results. In turn, that allows for cost reduction in hyperparameter studies. Here, we present first results on a concrete setup; we think that the understanding the mechanics of large system sizes is of fundamental interest for a further exploration of even larger models.

Figures

Figures reproduced from arXiv: 2506.16715 by the authors.

Figure 2
Figure 2. FIG. 2. EfficientNetB0...7 accuracy vs. noise intensity. Noise is [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Scaling of validation accuracy of EfficientNet type models [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. FIG. 4. Width of the cognition transition for EfficientNet type models [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

80 extracted references · 37 canonical work pages

  1. [1]

    author author K. R. \ Popper ,\ @noop title Logik der Forschung. Zur Erkenntnistheorie der modernen Naturwissenschaft \ ( publisher Julius Springer ,\ year 1935 ) NoStop

    work page 1935

  2. [2]

    Goodfellow , author Y

    author author I. Goodfellow , author Y. Bengio , \ and\ author A. Courville ,\ deeplearningbook.org title Deep Learning \ ( publisher MIT Press ,\ year 2016 ) NoStop

    work page 2016

  3. [3]

    Rombach , author A

    author author R. Rombach , author A. Blattmann , author D. Lorenz , author P. Esser , \ and\ author B. Ommer ,\ title title High-resolution image synthesis with latent diffusion models , \ in\ @noop booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognition \ ( year 2022 )\ pp.\ pages 10684--10695 NoStop

    work page 2022

  4. [4]

    Sohl-Dickstein , author E

    author author J. Sohl-Dickstein , author E. A. \ Weiss , author N. Maheswaranathan , \ and\ author S. Ganguli ,\ title title Deep unsupervised learning using nonequilibrium thermodynamics , \ @noop journal journal 32nd International Conference on Machine Learning, ICML 2015 \ volume 3 ( year 2015 ) NoStop

    work page 2015

  5. [5]

    Huang ,\ https://books.google.de/books?id=8bkBJgAACAAJ title Introduction to Statistical Physics, Second Edition \ ( publisher Taylor & Francis ,\ year 2009 ) NoStop

    author author K. Huang ,\ https://books.google.de/books?id=8bkBJgAACAAJ title Introduction to Statistical Physics, Second Edition \ ( publisher Taylor & Francis ,\ year 2009 ) NoStop

    work page 2009

  6. [6]

    Seidler \ and\ author M

    author author T. Seidler \ and\ author M. Abel ,\ title title Transition of model performance in dependence of the amount of data corruption with respect to network sizes , \ in\ 10.5220/0012435000003636 booktitle Proceedings of the 16th International Conference on Agents and Artificial Intelligence - Volume 3: ICAART ,\ organization INSTICC \ ( publisher...

    work page doi:10.5220/0012435000003636

  7. [7]

    Binder ,\ title title Finite size effects at phase transitions , \ in\ @noop booktitle Computational Methods in Field Theory ,\ editor edited by\ editor H

    author author K. Binder ,\ title title Finite size effects at phase transitions , \ in\ @noop booktitle Computational Methods in Field Theory ,\ editor edited by\ editor H. Gausterer \ and\ editor C. B. \ Lang \ ( publisher Springer Berlin Heidelberg ,\ address Berlin, Heidelberg ,\ year 1992 )\ pp.\ pages 59--125 NoStop

    work page 1992

  8. [8]

    Bahri , author J

    author author Y. Bahri , author J. Kadmon , author J. Pennington , author S. S. \ Schoenholz , author J. Sohl-Dickstein , \ and\ author S. Ganguli ,\ title title Statistical mechanics of deep learning , \ 10.1146/annurev-conmatphys-031119-050745 journal journal Annual Review of Condensed Matter Physics \ volume 11 ,\ pages 501--528 ( year 2020 ) NoStop

    doi:10.1146/annurev-conmatphys-031119-050745

Show all 80 references
  1. [10]

    Mézard \ and\ author A

    author author M. Mézard \ and\ author A. Montanari ,\ 10.1093/acprof:oso/9780198570837.001.0001 title Information, Physics, and Computation \ ( publisher Oxford University Press ,\ year 2009 ) NoStop

    2009

  2. [11]

    author author E. Gardner ,\ title title The space of interactions in neural network models , \ 10.1088/0305-4470/21/1/030 journal journal Journal of Physics A: Mathematical and General \ volume 21 ,\ pages 257 ( year 1988 ) NoStop

    1988 doi

  3. [12]

    Sompolinsky , author N

    author author H. Sompolinsky , author N. Tishby , \ and\ author H. S. \ Seung ,\ title title Learning from examples in large neural networks , \ 10.1103/PhysRevLett.65.1683 journal journal Phys. Rev. Lett. \ volume 65 ,\ pages 1683--1686 ( year 1990 ) NoStop

    doi

  4. [13]

    author author G. Györgyi ,\ title title First-order transition to perfect generalization in a neural network with binary synapses , \ 10.1103/PhysRevA.41.7097 journal journal Physical Review A \ volume 41 ,\ pages 7097--7100 ( year 1990 ) NoStop

    doi

  5. [15]

    author author T. L. \ Watkin , author A. Rau , \ and\ author M. Biehl ,\ title title The statistical mechanics of learning a rule , \ 10.1103/RevModPhys.65.499 journal journal Reviews of Modern Physics \ volume 65 ( year 1993 ),\ 10.1103/RevModPhys.65.499 NoStop

    doi

  6. [16]

    Biehl , author M

    author author M. Biehl , author M. Ahr , \ and\ author E. Schlösser ,\ title title Statistical physics of learning: Phase transitions in multilayered neural networks , \ 10.1007/bfb0108398 journal journal Advances in Solid State Physics 40 \ ,\ pages 819--826 ( year 2007 ) NoStop

    doi

  7. [17]

    Advani , author S

    author author M. Advani , author S. Lahiri , \ and\ author S. Ganguli ,\ title title Statistical mechanics of complex neural systems and high dimensional data , \ 10.1088/1742-5468/2013/03/P03014 journal journal Journal of Statistical Mechanics: Theory and Experiment \ volume ...

    2013 doi

  8. [18]

    Tian , author L

    author author C. Tian , author L. Fei , author W. Zheng , author Y. Xu , author W. Zuo , \ and\ author C.-W. \ Lin ,\ title title Deep learning on image denoising: An overview , \ https://doi.org/10.1016/j.neunet.2020.07.025 journal journal Neural Networks \ volume 131 ,\ page...

    2020 doi

  9. [19]

    Smilkov , author N

    author author D. Smilkov , author N. Thorat , author B. Kim , author F. B. \ Vi \' e gas , \ and\ author M. Wattenberg ,\ title title Smoothgrad: removing noise by adding noise , \ http://arxiv.org/abs/1706.03825 journal journal CoRR \ volume abs/1706.03825 ( year 2017 ) ,\ ht...

    2017 arXiv

  10. [20]

    Koziarski \ and\ author B

    author author M. Koziarski \ and\ author B. Cyganek ,\ title title Image recognition with deep neural networks in presence of noise – dealing with and taking advantage of distortions , \ 10.3233/ICA-170551 journal journal Integrated Computer-Aided Engineering \ volume 24 ,\ pa...

    doi

  11. [21]

    Drenkow , author N

    author author N. Drenkow , author N. Sani , author I. Shpitser , \ and\ author M. Unberath ,\ title title A systematic review of robustness in deep learning for computer vision: Mind the gap? \ @noop journal journal arXiv preprint arXiv:2112.00639 \ ( year 2021 ) NoStop

    2021 arXiv

  12. [22]

    Elgendy ,\ @noop title Deep Learning for Vision Systems \ ( publisher Mannings ,\ year 2020 ) NoStop

    author author M. Elgendy ,\ @noop title Deep Learning for Vision Systems \ ( publisher Mannings ,\ year 2020 ) NoStop

    2020

  13. [23]

    author author C. Boncelet ,\ title title Chapter 7 - Image Noise Models , \ in\ https://doi.org/10.1016/B978-0-12-374457-9.00007-X booktitle The Essential Guide to Image Processing ,\ editor edited by\ editor A. Bovik \ ( publisher Academic Press ,\ address Boston ,\ year 2009...

    doi

  14. [24]

    Ho , author A

    author author J. Ho , author A. Jain , \ and\ author P. Abbeel ,\ title title Denoising diffusion probabilistic models , \ @noop journal journal Advances in Neural Information Processing Systems \ ( year 2020 ) NoStop

    2020

  15. [25]

    James , author D

    author author G. James , author D. Witten , author T. Hastie , \ and\ author R. Tibshirani ,\ @noop title An Introduction to Statistical Learning: with Applications in R \ ( publisher Springer Publishing Company, Incorporated ,\ year 2014 ) NoStop

    2014

  16. [26]

    Heffern , author H

    author author E. Heffern , author H. Huelskamp , author S. Bahar , \ and\ author R. F. \ Inglis ,\ title title Phase transitions in biology: From bird flocks to population dynamics , \ 10.1098/rspb.2021.1111 journal journal Proceedings of the Royal Society B: Biological Scienc...

    2021

  17. [27]

    author author O. C. \ Martin , author R. Monasson , \ and\ author R. Zecchina ,\ title title Statistical mechanics methods and phase transitions in optimization problems , \ https://doi.org/10.1016/S0304-3975(01)00149-9 journal journal Theoretical Computer Science \ volume 265...

    doi

  18. [28]

    Perc , author J

    author author M. Perc , author J. J. \ Jordan , author D. G. \ Rand , author Z. Wang , author S. Boccaletti , \ and\ author A. Szolnoki ,\ title title Statistical physics of human cooperation , \ 10.1016/j.physrep.2017.05.004 journal journal Physics Reports \ volume 687 ( year...

    2017 doi

  19. [29]

    Rem , author N

    author author B. Rem , author N. Käming , author M. Tarnowski , author L. Asteria , author N. Fläschner , author C. Becker , author K. Sengstock , \ and\ author C. Weitenberg ,\ title title Identifying quantum phase transitions using artificial neural networks on experimental ...

    doi

  20. [30]

    Ni , author M

    author author Q. Ni , author M. Tang , author Y. Liu , \ and\ author Y.-C. \ Lai ,\ title title Machine learning dynamical phase transitions in complex networks , \ 10.1103/PhysRevE.100.052312 journal journal Phys. Rev. E \ volume 100 ,\ pages 052312 ( year 2019 ) NoStop

    doi

  21. [31]

    Carrasquilla \ and\ author R

    author author J. Carrasquilla \ and\ author R. G. \ Melko ,\ title title Machine learning phases of matter , \ 10.1038/nphys4035 journal journal Nature Physics \ volume 13 ,\ pages 431--434 ( year 2017 ) ,\ http://arxiv.org/abs/1605.01735 arXiv:1605.01735 [cond-mat.str-el] NoStop

    2017 arXiv

  22. [32]

    Canabarro , author F

    author author A. Canabarro , author F. F. \ Fanchini , author A. L. \ Malvezzi , author R. Pereira , \ and\ author R. Chaves ,\ title title Unveiling phase transitions with machine learning , \ 10.1103/PhysRevB.100.045129 journal journal Phys. Rev. B \ volume 100 ,\ pages 0451...

    doi

  23. [33]

    Giannetti , author B

    author author C. Giannetti , author B. Lucini , \ and\ author D. Vadacchino ,\ title title Machine learning as a universal tool for quantitative investigations of phase transitions , \ https://doi.org/10.1016/j.nuclphysb.2019.114639 journal journal Nuclear Physics B \ volume 9...

  24. [34]

    Suchsland \ and\ author S

    author author P. Suchsland \ and\ author S. Wessel ,\ title title Parameter diagnostics of phases and phase transition learning by neural networks , \ 10.1103/PhysRevB.97.174435 journal journal Phys. Rev. B \ volume 97 ,\ pages 174435 ( year 2018 ) NoStop

    doi

  25. [35]

    Cortes , author L

    author author C. Cortes , author L. D. \ Jackel , \ and\ author W.-P. \ Chiang ,\ title title Limits on learning machine accuracy imposed by data quality , \ @noop journal journal Advances in Neural Information Processing Systems \ volume 7 ( year 1994 ) NoStop

    1994

  26. [36]

    author author O. Y. \ Al-Jarrah , author P. D. \ Yoo , author S. Muhaidat , author G. K. \ Karagiannidis , \ and\ author K. Taha ,\ title title Efficient machine learning for big data: A review , \ https://doi.org/10.1016/j.bdr.2015.04.001 journal journal Big Data Research \ v...

    doi

  27. [37]

    Kaplan , author S

    author author J. Kaplan , author S. McCandlish , author T. Henighan , author T. B. \ Brown , author B. Chess , author R. Child , author S. Gray , author A. Radford , author J. Wu , \ and\ author D. Amodei ,\ title title Scaling laws for neural language models. \ https://arxiv....

    2001 arXiv

  28. [38]

    Deng , author W

    author author J. Deng , author W. Dong , author R. Socher , author L.-J. \ Li , author K. Li , \ and\ author L. Fei-Fei ,\ title title ImageNet: A Large-Scale Hierarchical Image Database , \ in\ @noop booktitle CVPR09 \ ( year 2009 ) NoStop

    2009

  29. [39]

    Krizhevsky , author G

    author author A. Krizhevsky , author G. Hinton , et al. ,\ title title Learning multiple layers of features from tiny images , \ @noop \ ( year 2009 ) NoStop

    2009

  30. [40]

    Tan \ and\ author Q

    author author M. Tan \ and\ author Q. V. \ Le ,\ https://arxiv.org/abs/1905.11946 title Efficientnet: Rethinking model scaling for convolutional neural networks , \ ( year 2020 ),\ http://arxiv.org/abs/1905.11946 arXiv:1905.11946 [cs.LG] NoStop

    1905 arXiv

  31. [41]

    van Kampen ,\ @noop title S tochastic P rocesses in P hysics and C hemistry \ ( publisher Elsevier Science Publishers, Amsterdam ,\ year 1992 ) NoStop

    author author N. van Kampen ,\ @noop title S tochastic P rocesses in P hysics and C hemistry \ ( publisher Elsevier Science Publishers, Amsterdam ,\ year 1992 ) NoStop

    1992

  32. [42]

    note For simplicity we write R ^ mn since R ^ mn is isomorphic to R ^ m n . Stop

  33. [43]

    Meyn \ and\ author R

    author author S. Meyn \ and\ author R. Tweedie ,\ title title Markov chains and stochastic stability , \ 10.2307/2965732 journal journal Journal of the American Statistical Association \ volume 92 ( year 1993 ),\ 10.2307/2965732 NoStop

    1993 doi

  34. [44]

    Chollet et al

    author author F. Chollet et al. ,\ @noop title Keras , \ howpublished https://keras.io ( year 2015 ) NoStop

    2015

  35. [45]

    Howard ,\ https://github.com/fastai/imagenette title Imagenette: A smaller subset of 10 easily classified classes from imagenet , \ ( year 2019 ) NoStop

    author author J. Howard ,\ https://github.com/fastai/imagenette title Imagenette: A smaller subset of 10 easily classified classes from imagenet , \ ( year 2019 ) NoStop

    2019

  36. [46]

    Stanley , author L

    author author H. Stanley , author L. Amaral , author P. Gopikrishnan , author P. Ivanov , author T. Keitt , \ and\ author V. Plerou ,\ title title Scale invariance and universality: organizing principles in complex systems , \ https://doi.org/10.1016/S0378-4371(00)00195-3 jour...

    doi

  37. [47]

    Lovejoy ,\ ed.,\ title Scaling and scale invariance , \ pp.\ pages 1244--1256 ,\ in\ Sagar2023 ( year 2023 ) NoStop

    editor S. Lovejoy ,\ ed.,\ title Scaling and scale invariance , \ pp.\ pages 1244--1256 ,\ in\ Sagar2023 ( year 2023 ) NoStop

    2023

  38. [48]

    author author L. C. \ Evans \ and\ author J. Spruck ,\ title title Motion of level sets by mean curvature. i , \ 10.4310/jdg/1214446559 journal journal Journal of Differential Geometry \ volume 33 ( year 1991 ),\ 10.4310/jdg/1214446559 NoStop

    1991

  39. [49]

    author author L. C. \ Evans , author H. M. \ Soner , \ and\ author P. E. \ Souganidis ,\ title title Phase transitions and generalized motion by mean curvature , \ 10.1002/cpa.3160450903 journal journal Communications on Pure and Applied Mathematics \ volume 45 ,\ pages 1097--...

    doi

  40. [50]

    Gardiner ,\ @noop title Handbook of stochastic methods for physics, chemistry, and the natural sciences \ ( publisher Springer ,\ year 1985 ) NoStop

    author author C. Gardiner ,\ @noop title Handbook of stochastic methods for physics, chemistry, and the natural sciences \ ( publisher Springer ,\ year 1985 ) NoStop

    1985

  41. [51]

    Carleo , author I

    author author G. Carleo , author I. Cirac , author K. Cranmer , author L. Daudet , author M. Schuld , author N. Tishby , author L. Vogt-Maranto , \ and\ author L. Zdeborová ,\ title title Machine learning and the physical sciences , \ 10.1103/RevModPhys.91.045002 journal journ...

    doi

  42. [52]

    author author H. S. \ Seung , author H. Sompolinsky , \ and\ author N. Tishby ,\ title title Statistical mechanics of learning from examples , \ 10.1103/PhysRevA.45.6056 journal journal Physical Review A \ volume 45 ,\ pages 6056--6091 ( year 1992 b ) NoStop

    doi

  43. [53]

    Xiao , author K

    author author H. Xiao , author K. Rasul , \ and\ author R. Vollgraf ,\ title title Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms , \ http://arxiv.org/abs/1708.07747 journal journal arXiv preprint arXiv:1708.07747 \ ( year 2017 ) NoStop

    2017 arXiv

  44. [54]

    Ising ,\ title title Beitrag zur theorie des ferromagnetismus , \ 10.1007/BF02980577 journal journal Zeitschrift für Physik \ volume 31 ,\ pages 253--258 ( year 1925 ) NoStop

    author author E. Ising ,\ title title Beitrag zur theorie des ferromagnetismus , \ 10.1007/BF02980577 journal journal Zeitschrift für Physik \ volume 31 ,\ pages 253--258 ( year 1925 ) NoStop

    1925 doi

  45. [55]

    author author C. E. \ Shannon ,\ title title A mathematical theory of communication , \ 10.1002/j.1538-7305.1948.tb01338.x journal journal The Bell System Technical Journal \ volume 27 ,\ pages 379--423 ( year 1948 ) NoStop

    1948

  46. [56]

    author author A. Lesne ,\ title title Shannon entropy: A rigorous notion at the crossroads between probability, information theory, dynamical systems and statistical physics , \ 10.1017/S0960129512000783 journal journal Mathematical Structures in Computer Science \ volume 24 (...

    doi

  47. [57]

    Quade , author M

    author author M. Quade , author M. Abel , author J. N. \ Kutz , \ and\ author S. L. \ Brunton ,\ title title Sparse identification of nonlinear dynamics for rapid model recovery , \ 10.1063/1.5027470 journal journal Chaos \ volume 28 ( year 2018 ),\ 10.1063/1.5027470 NoStop

    doi

  48. [58]

    LeCun , author C

    author author Y. LeCun , author C. Cortes , \ and\ author C. J. \ Burges ,\ title title MNIST handwritten digit database , \ @noop journal journal ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist \ volume 2 ( year 2010 ) NoStop

    2010

  49. [59]

    author author L. Deng ,\ title title The mnist database of handwritten digit images for machine learning research , \ 10.1109/MSP.2012.2211477 journal journal IEEE Signal Processing Magazine \ volume 29 ,\ pages 141--142 ( year 2012 ) NoStop

    2012

  50. [60]

    Cohen , author S

    author author G. Cohen , author S. Afshar , author J. Tapson , \ and\ author A. V. \ Schaik ,\ title title EMNIST: Extending MNIST to handwritten letters , \ 10.1109/IJCNN.2017.7966217 journal journal Proceedings of the International Joint Conference on Neural Networks \ volum...

    2017

  51. [61]

    Quade , author M

    author author M. Quade , author M. Abel , author K. Shafi , author R. K. \ Niven , \ and\ author B. R. \ Noack ,\ title title Prediction of dynamical systems by symbolic regression , \ 10.1103/PhysRevE.94.012214 journal journal Physical Review E \ volume 94 ( year 2016 ),\ 10....

    doi

  52. [62]

    Roscher , author B

    author author R. Roscher , author B. Bohn , author M. F. \ Duarte , \ and\ author J. Garcke ,\ title title Explainable machine learning for scientific insights and discoveries , \ 10.1109/ACCESS.2020.2976199 journal journal IEEE Access \ volume 8 ( year 2020 ),\ 10.1109/ACCESS...

    2020

  53. [63]

    author author F. Rosenblatt ,\ title title The perceptron: A probabilistic model for information storage and organization in the brain , \ 10.1037/h0042519 journal journal Psychological Review \ volume 65 ( year 1958 ),\ 10.1037/h0042519 NoStop

    1958 doi

  54. [64]

    author author J. R. \ Koza ,\ title title Genetic programming as a means for programming computers by natural selection , \ 10.1007/BF00175355 journal journal Statistics and Computing \ volume 4 ( year 1994 ),\ 10.1007/BF00175355 NoStop

    1994 doi

  55. [65]

    author author L. G. \ Valiant ,\ title title A theory of the learnable , \ 10.1145/1968.1972 journal journal Commun. ACM \ volume 27 ,\ pages 1134–1142 ( year 1984 ) NoStop

    1968

  56. [66]

    author author J. Hopfield ,\ title title Neural networks and physical systems with emergent collective computational abilities , \ 10.1073/pnas.79.8.2554 journal journal Proceedings of the National Academy of Sciences of the United States of America \ volume 79 ,\ pages 2554--...

    doi

  57. [67]

    Gardner ,\ title title Maximum storage capacity in neural networks , \ 10.1209/0295-5075/4/4/016 journal journal Europhysics Letters \ volume 4 ,\ pages 481 ( year 1987 ) NoStop

    author author E. Gardner ,\ title title Maximum storage capacity in neural networks , \ 10.1209/0295-5075/4/4/016 journal journal Europhysics Letters \ volume 4 ,\ pages 481 ( year 1987 ) NoStop

    1987 doi

  58. [68]

    Perc ,\ 10.1016/j.physleta.2016.06.017 title Phase transitions in models of human cooperation , \ ( year 2016 ) NoStop

    author author M. Perc ,\ 10.1016/j.physleta.2016.06.017 title Phase transitions in models of human cooperation , \ ( year 2016 ) NoStop

    2016 doi

  59. [69]

    Baldominos , author Y

    author author A. Baldominos , author Y. Saez , \ and\ author P. Isasi ,\ title title A survey of handwritten character recognition with MNIST and EMNIST , \ 10.3390/app9153169 journal journal Applied Sciences (Switzerland) \ volume 9 ( year 2019 ),\ 10.3390/app9153169 NoStop

    doi

  60. [70]

    Pedregosa , author G

    author author F. Pedregosa , author G. Varoquaux , author A. Gramfort , author V. Michel , author B. Thirion , author O. Grisel , author M. Blondel , author P. Prettenhofer , author R. Weiss , author V. Dubourg , author J. Vanderplas , author A. Passos , author D. Cournapeau ,...

    2011

  61. [71]

    author author Tensorflow Development Team ,\ @noop title TensorFlow Datasets , a collection of ready-to-use datasets , \ howpublished https://www.tensorflow.org/datasets ( year 2023 ),\ note accessed: 13.11.2023 NoStop

    2023

  62. [72]

    author author Dask Development Team ,\ https://dask.org title Dask: Library for dynamic task scheduling , \ ( year 2016 ) NoStop

    2016

  63. [73]

    Seidler , author F

    author author T. Seidler , author F. Emmerich , author K. Ehlert , author R. Berner , author O. Nagel-Kanzler , author N. Schultz , author M. Quade , author M. G. \ Schultz , \ and\ author M. Abel ,\ title title Mantik: A workflow platform for the development of artificial int...

    2023

  64. [74]

    author author C. M. \ Bishop ,\ @noop title Pattern Recognition and Machine Learning (Information Science and Statistics) \ ( publisher Springer-Verlag ,\ address Berlin, Heidelberg ,\ year 2006 ) NoStop

    2006

  65. [75]

    Lecun , author L

    author author Y. Lecun , author L. Bottou , author Y. Bengio , \ and\ author P. Haffner ,\ title title Gradient-based learning applied to document recognition , \ 10.1109/5.726791 journal journal Proceedings of the IEEE \ volume 86 ,\ pages 2278--2324 ( year 1998 ) NoStop

    doi

  66. [76]

    Szegedy , author W

    author author C. Szegedy , author W. Zaremba , author I. Sutskever , author J. Bruna , author D. Erhan , author I. J. \ Goodfellow , \ and\ author R. Fergus ,\ title title Intriguing properties of neural networks , \ in\ http://arxiv.org/abs/1312.6199 booktitle 2nd Internation...

    2014 arXiv

  67. [77]

    author author I. J. \ Goodfellow , author J. Shlens , \ and\ author C. Szegedy ,\ title title Explaining and harnessing adversarial examples , \ in\ http://arxiv.org/abs/1412.6572 booktitle 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA...

    2015 arXiv

  68. [78]

    Nishimori \ and\ author G

    author author H. Nishimori \ and\ author G. Ortiz ,\ 10.1093/acprof:oso/9780199577224.001.0001 title Elements of Phase Transitions and Critical Phenomena \ ( publisher Oxford University Press ,\ year 2010 )\ http://arxiv.org/abs/https://academic.oup.com/book/8876/book-pdf/5343...

    2010

  69. [79]

    Michaelis , author B

    author author C. Michaelis , author B. Mitzkus , author R. Geirhos , author E. Rusak , author O. Bringmann , author A. S. \ Ecker , author M. Bethge , \ and\ author W. Brendel ,\ title title Benchmarking robustness in object detection: Autonomous driving when winter is coming ...

    1907 arXiv

  70. [80]

    Guo , author U

    author author J. Guo , author U. Kurup , \ and\ author M. Shah ,\ title title Is it safe to drive? an overview of factors, metrics, and datasets for driveability assessment in autonomous driving , \ 10.1109/TITS.2019.2926042 journal journal IEEE Transactions on Intelligent Tra...

    2019

  71. [81]

    Koziarski \ and\ author B

    author author M. Koziarski \ and\ author B. Cyganek ,\ title title Deep neural image denoising , \ \ ( year 2016 )\ pp.\ pages 163--173 NoStop

    2016

  72. [82]

    editor B. S. \ Daya Sagar , editor Q. Cheng , editor J. McKinley , \ and\ editor F. Agterberg ,\ eds.,\ 10.1007/978-3-030-85040-1_282 title Encyclopedia of Mathematical Geosciences \ ( publisher Springer International Publishing ,\ year 2023 ) NoStop

    doi

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.