pith. sign in

arxiv: 2606.01841 · v1 · pith:JBO3MFWMnew · submitted 2026-06-01 · 🧬 q-bio.NC

The Neuromorphic Supremacy

Yuliya Tsybina , Ivan Y. Tyukin , Alexander N. Gorban , Victor Kazantsev , Dianhui Wang , Susanna Gordleeva This is my paper

Pith reviewed 2026-06-28 11:59 UTC · model grok-4.3

classification 🧬 q-bio.NC
keywords neuromorphic circuitshybrid neural networksfew-shot learningnoise robustnessspiking dynamicsastrocytic modulationneuromorphic supremacy
0
0 comments X

The pith

Hybrid neural networks embedding neuromorphic circuits learn accurately from few examples and resist noise where standard models fail.

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

The paper shows that adding genuine neuromorphic circuits with astrocytic modulation and spiking dynamics to standard artificial neural networks produces hybrid models that reach high accuracy using only a few training examples per class. These models also maintain performance when faced with occlusion and impulse noise that break conventional deep learning approaches. This difference, called neuromorphic supremacy, matters because it offers a route to AI systems that function like biological ones in data-poor and noisy real-world conditions. A reader would care if they want AI that adapts with less data and more reliability.

Core claim

Embedding novel genuine neuromorphic circuits into conventional artificial neural network architectures allows the resulting hybrid models to achieve high accuracy from few training examples per class and to sustain high performance under occlusion and impulse noise that cause performance collapse in standard models without neuromorphic adaptation, a regime termed neuromorphic supremacy.

What carries the argument

Genuine neuromorphic circuits with astrocytic modulation and spiking dynamics integrated into ANN architectures.

If this is right

  • Hybrid models achieve high accuracy from few training examples per class across standard benchmarks.
  • They sustain high performance under occlusion and impulse noise.
  • Standard models without the adaptation suffer performance collapse in those conditions.
  • This points toward a principled foundation for perception in embodied AI systems operating in noisy, data-scarce environments.

Where Pith is reading between the lines

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

  • Integration of these circuits could allow AI training with smaller datasets in practical deployments.
  • Further biological features might be added to enhance robustness in additional ways.
  • Deployment in physical robots would test if the supremacy holds beyond simulated benchmarks.

Load-bearing premise

The neuromorphic circuits integrate into existing ANN training pipelines without causing instabilities or requiring unreported dataset-specific adjustments.

What would settle it

An experiment showing that the hybrid models do not achieve higher few-shot accuracy or better noise robustness than standard models on the benchmarks used.

read the original abstract

Live neural systems demonstrate remarkable capabilities to learn new behavior and patterns from mere few examples and are known to operate robustly under severe sensory noise. These capabilities, however, remain largely out of reach for modern artificial neural networks, including deep learning models. We show that this gap can be bridged by embedding novel genuine neuromorphic circuits into conventional artificial neural network architectures. These circuits comprise astrocytic modulation and spiking dynamics inherent to biological neural structures. Tested across standard benchmarks representing tasks of varying complexity, the hybrid models achieve high accuracy from few training examples per class and sustain high performance under occlusion and impulse noise that cause performance collapse in standard models without neuromorphic adaptation. We term this phenomenon neuromorphic supremacy - a regime in which architectures grounded in neurobiology decisively outperform classical deep learning, pointing toward a principled foundation for perception in embodied AI systems operating in noisy, data-scarce environments.

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

0 major / 2 minor

Summary. The paper claims that embedding novel neuromorphic circuits (comprising astrocytic modulation and spiking dynamics) into standard ANN architectures bridges the gap between biological learning robustness and artificial networks. On standard benchmarks of varying complexity, the resulting hybrid models achieve high accuracy from few training examples per class and maintain performance under occlusion and impulse noise that degrade conventional models, a regime termed 'neuromorphic supremacy'.

Significance. If the reported empirical comparisons hold under the described conditions, the work would demonstrate a concrete, biologically grounded route to improving few-shot generalization and noise robustness in ANNs, with potential implications for embodied AI in data-scarce environments.

minor comments (2)
  1. [Abstract] Abstract: the claim of empirical superiority is stated without any numerical accuracies, training-set sizes, statistical tests, or baseline comparisons. Although the full manuscript reportedly supplies direct comparisons, the abstract should include at least one or two key quantitative results so that the central claim can be evaluated at a glance.
  2. The term 'neuromorphic supremacy' is introduced without a precise operational definition (e.g., a threshold performance gap or a specific set of conditions). A short clarifying sentence would prevent ambiguity.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of the manuscript, recognition of its potential significance for embodied AI, and recommendation for minor revision. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper presents an empirical claim that hybrid neuromorphic-ANN models outperform standard deep learning on few-shot and noisy benchmarks. No equations, derivations, fitted parameters renamed as predictions, or self-citation chains appear in the abstract or described structure. The term 'neuromorphic supremacy' is introduced as a label for the observed performance gap rather than derived from prior self-referential results. The central results rest on direct experimental comparisons that remain externally falsifiable and do not reduce to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Abstract supplies insufficient technical detail to enumerate free parameters, background axioms, or new entities beyond the high-level description of the circuits themselves.

invented entities (1)
  • neuromorphic circuits comprising astrocytic modulation no independent evidence
    purpose: To confer few-shot learning and noise robustness on conventional ANNs
    Presented as novel components in the abstract; no independent evidence or falsifiable prediction supplied.

pith-pipeline@v0.9.1-grok · 5692 in / 1053 out tokens · 31039 ms · 2026-06-28T11:59:50.664272+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

76 extracted references · 26 canonical work pages

  1. [1]

    The perceptron: a probabilistic model for information storage and organization in the brain.Psychological review65, 386 (1958)

    Rosenblatt, F. The perceptron: a probabilistic model for information storage and organization in the brain.Psychological review65, 386 (1958)

    1958

  2. [2]

    & Bengio, Y

    LeCun, Y. & Bengio, Y. Convolutional networks for images, speech, and time series.The handbook of brain theory and neural networks(1998)

    1998

  3. [3]

    Vaswani, A.et al.Attention is all you need.Advances in neural information processing systems30(2017)

    2017

  4. [4]

    Hernstr¨ om, V.et al.Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): a randomised, controlled, parallel-group, non-inferiority, single- blinded, screening accuracy study.The Lancet Digital Health7, e175–e183 (2025)

    2025

  5. [5]

    Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad

    Google DeepMind. Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad. https: //deepmind.google/discover/blog/advanced-version-of-gemini-with-deep-think -officially-achieves-gold-medal-standard-at-the-international-mathematical-oly mpiad/ (2025). Accessed: 2026-04-26

    2025

  6. [6]

    Abramson, J.et al.Accurate structure prediction of biomolecular interactions with AlphaFold 3.Nature630, 493–500 (2024). 27

    2024

  7. [7]

    Merchant, A.et al.Scaling deep learning for materials discovery.Nature624, 80–85 (2023)

    2023

  8. [8]

    for computational protein design

    The Royal Swedish Academy of Sciences. The Nobel Prize in Chemistry 2024. Press release, NobelPrize.org (2024). Awarded one half to David Baker “for computational protein design” and the other half jointly to Demis Hassabis and John Jumper “for protein structure prediction”. Available at https://www.nobe lprize.org/prizes/chemistry/2024/press-release/. Ac...

    2024

  9. [9]

    & Higham, D

    Bastounis, A., Tyukin, I. & Higham, D. A survey of inevitability results in ai instability.Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences(2026)

    2026

  10. [10]

    Bastounis, A.et al.The boundaries of verifiable accuracy, robustness, and generalisation in deep learning (2023)

    2023

  11. [11]

    arXiv:2203.15556

    Hoffmann, J.et al.Training compute-optimal large language models (2022). arXiv:2203.15556

    Pith/arXiv arXiv 2022

  12. [12]

    zero-shot

    Udandarao, V.et al.No “zero-shot” without exponential data: Pretrain- ing concept frequency determines multimodal model performance (2024). arXiv:2404.04125

    arXiv 2024

  13. [13]

    Wood, J. N. Newborn chickens generate invariant object representations at the onset of visual object experience.Proceedings of the National Academy of Sciences 110, 14000–14005 (2013)

    2013

  14. [14]

    Wood, J. N. & Wood, S. M. W. One-shot learning of view-invariant object representations in newborn chicks.Cognition199, 104192 (2020)

    2020

  15. [15]

    arXiv:1808.08750

    Geirhos, R.et al.Generalisation in humans and deep neural networks (2018). arXiv:1808.08750

    arXiv 2018

  16. [16]

    Geirhos, R.et al.Shortcut learning in deep neural networks.Nature Machine Intelligence2, 665–673 (2020)

    2020

  17. [17]

    Hennessy, J. L. & Patterson, D. A. A new golden age for computer architecture. Communications of the ACM62, 48–60 (2019)

    2019

  18. [18]

    & Brown, T

    Hernandez, D. & Brown, T. B. Measuring the algorithmic efficiency of neural networks.arXiv preprint arXiv:2005.04305(2020)

    arXiv 2005

  19. [19]

    National Research Council.The Future of Computing Performance: Game Over or Next Level?(The National Academies Press, Washington, DC, 2011)

    2011

  20. [20]

    & Thompson, N

    Sherry, Y. & Thompson, N. C. How fast do algorithms improve?Proceedings of the IEEE109, 1768–1777 (2021). 28

    2021

  21. [21]

    Felleman, D. J. & Van Essen, D. C. Distributed hierarchical processing in the primate cerebral cortex.Cerebral Cortex1, 1–47 (1991)

    1991

  22. [22]

    Hubel, D. H. & Wiesel, T. N. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex.The Journal of Physiology160, 106–154 (1962). URL http://dx.doi.org/10.1113/jphysiol.1962.sp006837

    work page doi:10.1113/jphysiol.1962.sp006837 1962

  23. [23]

    & Koch, C

    Itti, L. & Koch, C. Computational modelling of visual attention.Nature Reviews Neuroscience2, 194–203 (2001). URL http://dx.doi.org/10.1038/35058500

    work page doi:10.1038/35058500 2001

  24. [24]

    URL http://dx.doi.org/10.10 16/j.neunet.2019.09.036

    Taherkhani, A.et al.A review of learning in biologically plausible spiking neural networks.Neural Networks122, 253–272 (2020). URL http://dx.doi.org/10.10 16/j.neunet.2019.09.036

    2020

  25. [25]

    & Panda, P

    Roy, K., Jaiswal, A. & Panda, P. Towards spike-based machine intelligence with neuromorphic computing.Nature575, 607–617 (2019). URL http://dx.doi.org /10.1038/s41586-019-1677-2

    work page doi:10.1038/s41586-019-1677-2 2019

  26. [26]

    & Araque, A

    Perea, G. & Araque, A. Astrocytes potentiate transmitter release at single hippocampal synapses.Science317, 1083–1086 (2007)

    2007

  27. [27]

    URL http://dx.doi.org/10.1109/TNNLS.2023.3335450

    Gordleeva, S.et al.Situation-based neuromorphic memory in spiking neuron- astrocyte network.IEEE Transactions on Neural Networks and Learning Systems 36, 881–895 (2023). URL http://dx.doi.org/10.1109/TNNLS.2023.3335450

    work page doi:10.1109/tnnls.2023.3335450 2023

  28. [28]

    & Volterra, A

    Santello, M., Toni, N. & Volterra, A. Astrocyte function from information pro- cessing to cognition and cognitive impairment.Nature Neuroscience22, 154–166 (2019). URL http://dx.doi.org/10.1038/s41593-018-0325-8

    work page doi:10.1038/s41593-018-0325-8 2019

  29. [29]

    & Papouin, T

    Murphy-Royal, C., Ching, S. & Papouin, T. A conceptual framework for astrocyte function.Nature Neuroscience26, 1848–1856 (2023). URL http://dx.doi.org/1 0.1038/s41593-023-01448-8

    2023

  30. [30]

    & Schmidhuber, J

    Hochreiter, S. & Schmidhuber, J. Long short-term memory.Neural computation 9, 1735–1780 (1997)

    1997

  31. [31]

    Hopfield, J. J. Neural networks and physical systems with emergent collective computational abilities.Proceedings of the National Academy of Sciences79, 2554–2558 (1982). URL http://dx.doi.org/10.1073/pnas.79.8.2554

    work page doi:10.1073/pnas.79.8.2554 1982

  32. [32]

    & Vollgraf, R

    Xiao, H., Rasul, K. & Vollgraf, R. Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms.arXiv preprint arXiv:1708.07747 (2017)

    Pith/arXiv arXiv 2017

  33. [33]

    M., Salakhutdinov, R

    Lake, B. M., Salakhutdinov, R. & Tenenbaum, J. B. Human-level concept learning through probabilistic program induction.Science350, 1332–1338 (2015). 29

    2015

  34. [34]

    Quantum computing and the entanglement frontier (2012)

    Preskill, J. Quantum computing and the entanglement frontier (2012). arXiv:1203.5813

    Pith/arXiv arXiv 2012

  35. [35]

    Elman, J. L. Finding structure in time.Cognitive Science14, 179–211 (1990). URL http://dx.doi.org/10.1207/s15516709cog1402 1

    work page doi:10.1207/s15516709cog1402 1990

  36. [36]

    Bidirectional associative memories.IEEE Transactions on Systems, Man, and Cybernetics18, 49–60 (1988)

    Kosko, B. Bidirectional associative memories.IEEE Transactions on Systems, Man, and Cybernetics18, 49–60 (1988). URL http://dx.doi.org/10.1109/21.87 054

    work page doi:10.1109/21.87 1988

  37. [37]

    & Xiao, J

    Zhao, Y. & Xiao, J. An adiabatic method to train binarized artificial neural networks.Scientific Reports11(2021). URL http://dx.doi.org/10.1038/s4159 8-021-99191-2

    work page doi:10.1038/s4159 2021

  38. [38]

    arXiv.org , author=

    Qin, H.et al.Binary neural networks: A survey.Pattern Recognition105, 107281 (2020). URL http://dx.doi.org/10.1016/j.patcog.2020.107281

    work page doi:10.1016/j.patcog.2020.107281 2020

  39. [39]

    Lake, Ruslan Salakhutdinov, and Joshua B

    Lake, B. M., Salakhutdinov, R. & Tenenbaum, J. B. Human-level concept learning through probabilistic program induction.Science350, 1332–1338 (2015). URL http://dx.doi.org/10.1126/science.aab3050

    work page doi:10.1126/science.aab3050 2015

  40. [40]

    Zeiler, M. D. & Fergus, R.Visualizing and Understanding Convolutional Net- works, 818–833 (Springer International Publishing, 2014). URL http://dx.doi.o rg/10.1007/978-3-319-10590-1 53

    work page doi:10.1007/978-3-319-10590-1 2014

  41. [41]

    & Krotov, D

    Kozachkov, L., Slotine, J.-J. & Krotov, D. Neuron–astrocyte associative memory. Proceedings of the National Academy of Sciences122(2025). URL http://dx.d oi.org/10.1073/pnas.2417788122

    work page doi:10.1073/pnas.2417788122 2025

  42. [42]

    Kozachkov, L., Kastanenka, K. V. & Krotov, D. Building transformers from neurons and astrocytes.Proceedings of the National Academy of Sciences120 (2023). URL http://dx.doi.org/10.1073/pnas.2219150120

    work page doi:10.1073/pnas.2219150120 2023

  43. [43]

    URL http://dx.doi.org /10.1038/s41467-017-00248-6

    Zuo, F.et al.Habituation based synaptic plasticity and organismic learning in a quantum perovskite.Nature Communications8(2017). URL http://dx.doi.org /10.1038/s41467-017-00248-6

    work page doi:10.1038/s41467-017-00248-6 2017

  44. [44]

    R., Ganjtabesh, M., Thorpe, S

    Kheradpisheh, S. R., Ganjtabesh, M., Thorpe, S. J. & Masquelier, T. Stdp- based spiking deep convolutional neural networks for object recognition.Neural Networks99, 56–67 (2018). URL http://dx.doi.org/10.1016/j.neunet.2017.12.00 5

    work page doi:10.1016/j.neunet.2017.12.00 2018

  45. [45]

    & Thorpe, S

    Masquelier, T. & Thorpe, S. J. Unsupervised learning of visual features through spike timing dependent plasticity.PLoS Computational Biology3, e31 (2007). URL http://dx.doi.org/10.1371/journal.pcbi.0030031. 30

    work page doi:10.1371/journal.pcbi.0030031 2007

  46. [46]

    Rao, R. P. N. & Sejnowski, T. J. Spike-timing-dependent hebbian plasticity as temporal difference learning.Neural Computation13, 2221–2237 (2001). URL http://dx.doi.org/10.1162/089976601750541787

    work page doi:10.1162/089976601750541787 2001

  47. [47]

    & Basu, A

    Roy, S. & Basu, A. An online unsupervised structural plasticity algorithm for spiking neural networks.IEEE Transactions on Neural Networks and Learning Systems28, 900–910 (2017). URL http://dx.doi.org/10.1109/TNNLS.2016.2582 517

    work page doi:10.1109/tnnls.2016.2582 2017

  48. [48]

    URL http://dx.doi.org/10.1142/97818 48162778 0008

    Maass, W.Liquid State Machines: Motivation, Theory, and Applications, 275–296 (IMPERIAL COLLEGE PRESS, 2011). URL http://dx.doi.org/10.1142/97818 48162778 0008

    work page doi:10.1142/97818 2011

  49. [49]

    & Campenhout, J

    Schrauwen, B., D’Haene, M., Verstraeten, D. & Campenhout, J. V. Compact hardware liquid state machines on fpga for real-time speech recognition.Neural Networks21, 511–523 (2008). URL http://dx.doi.org/10.1016/j.neunet.2007.12. 009

    work page doi:10.1016/j.neunet.2007.12 2008

  50. [50]

    Simple model of spiking neurons.IEEE Transactions on Neural Networks14, 1569–1572 (2003)

    Izhikevich, E. Simple model of spiking neurons.IEEE Transactions on Neural Networks14, 1569–1572 (2003)

    2003

  51. [51]

    Kazantsev, V. B. & Asatryan, S. Y. Bistability induces episodic spike com- munication by inhibitory neurons in neuronal networks.Physical Review E84 (2011)

    2011

  52. [52]

    M., Gordleeva, S

    Esir, P. M., Gordleeva, S. Y., Simonov, A. Y., Pisarchik, A. N. & Kazantsev, V. B. Conduction delays can enhance formation of up and down states in spiking neuronal networks.Physical Review E98(2018)

    2018

  53. [53]

    & Cornell-Bell, A

    Ullah, G., Jung, P. & Cornell-Bell, A. Anti-phase calcium oscillations in astrocytes via inositol (1, 4, 5)-trisphosphate regeneration.Cell Calcium39, 197–208 (2006)

    2006

  54. [54]

    Yamamoto, T., Ochalski, A., Hertzberg, E. L. & Nagy, J. I. On the organization of astrocytic gap junctions in rat brain as suggested by LM and EM immuno- histochemistry of connexin43 expression.The Journal of Comparative Neurology 302, 853–883 (1990)

    1990

  55. [55]

    Nagy, J. I. & Rash, J. E. Connexins and gap junctions of astrocytes and oligodendrocytes in the CNS.Brain Research Reviews32, 29–44 (2000)

    2000

  56. [56]

    M., Fellin, T., Takano, H., Dong, J.-H

    Halassa, M. M., Fellin, T., Takano, H., Dong, J.-H. & Haydon, P. G. Synaptic islands defined by the territory of a single astrocyte.Journal of Neuroscience27, 6473–6477 (2007)

    2007

  57. [57]

    Y., Stasenko, S

    Gordleeva, S. Y., Stasenko, S. V., Semyanov, A. V., Dityatev, A. E. & Kazantsev, V. B. Bi-directional astrocytic regulation of neuronal activity within a network. Frontiers in Computational Neuroscience6(2012). 31

    2012

  58. [58]

    V.et al.Neuronal synchronization enhanced by neuron–astrocyte interaction.Nonlinear Dynamics97, 647–662 (2019)

    Pankratova, E. V.et al.Neuronal synchronization enhanced by neuron–astrocyte interaction.Nonlinear Dynamics97, 647–662 (2019)

    2019

  59. [59]

    & Araque, A

    Navarrete, M. & Araque, A. Endocannabinoids mediate neuron-astrocyte communication.Neuron57, 883–893 (2008)

    2008

  60. [60]

    & Araque, A

    Navarrete, M. & Araque, A. Endocannabinoids potentiate synaptic transmission through stimulation of astrocytes.Neuron68, 113–126 (2010)

    2010

  61. [61]

    Lecun, L

    Lecun, Y., Bottou, L., Bengio, Y. & Haffner, P. Gradient-based learning applied to document recognition.Proceedings of the IEEE86, 2278–2324 (1998). URL http://dx.doi.org/10.1109/5.726791

    work page doi:10.1109/5.726791 1998

  62. [62]

    & Szegedy, C

    Ioffe, S. & Szegedy, C. Batch normalization: Accelerating deep network training by reducing internal covariate shift (2015). URL https://arxiv.org/abs/1502.031 67

    2015

  63. [63]

    Nair V., H. G. E. Rectified linear units improve restricted boltzmann machines. ICML(2010)

    2010

  64. [64]

    & Hinton, G

    Sutskever, I., Martens, J., Dahl, G. & Hinton, G. Dasgupta, S. & McAllester, D. (eds)On the importance of initialization and momentum in deep learning. (eds Dasgupta, S. & McAllester, D.)Proceedings of the 30th International Conference on Machine Learning, Vol. 28 ofProceedings of Machine Learning Research, 1139– 1147 (PMLR, Atlanta, Georgia, USA, 2013). ...

    2013

  65. [65]

    Hin- ton,Imagenet classification with deep convolutional neu- ral networks, Commun

    Krizhevsky, A., Sutskever, I. & Hinton, G. E. Imagenet classification with deep convolutional neural networks.Communications of the ACM60, 84–90 (2012). URL http://dx.doi.org/10.1145/3065386

    work page doi:10.1145/3065386 2012

  66. [66]

    & Bengio, Y

    Glorot, X. & Bengio, Y. Teh, Y. W. & Titterington, M. (eds)Understand- ing the difficulty of training deep feedforward neural networks. (eds Teh, Y. W. & Titterington, M.)Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, Vol. 9 ofProceedings of Machine Learning Research, 249–256 (PMLR, Chia Laguna Resort, Sa...

    2010

  67. [67]

    URL https://arxiv.org/abs/2010.11929

    Dosovitskiy, A.et al.An image is worth 16x16 words: Transformers for image recognition at scale (2020). URL https://arxiv.org/abs/2010.11929

    Pith/arXiv arXiv 2020

  68. [68]

    BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding

    Jacob, D., Ming-Wei, C., Kenton, L. & Kristina, T. Bert: Pre-training of deep bidirectional transformers for language understanding (2019). URL http://dx.d oi.org/10.18653/v1/N19-1423

    work page doi:10.18653/v1/n19-1423 2019

  69. [69]

    Vaswani, A.et al.Attention is all you need.Advances in neural information processing systems30(2017). 32

    2017

  70. [70]

    Kingma, D. P. & Ba, J. Adam: A method for stochastic optimization (2014). URL https://arxiv.org/abs/1412.6980

    Pith/arXiv arXiv 2014

  71. [71]

    & Sun, J

    He, K., Zhang, X., Ren, S. & Sun, J. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification (2015)

    2015

  72. [72]

    & Mosin, S.An Effect of Binarization on Handwritten Digits Recognition by Hierarchical Neural Networks, 94–106 (Springer Interna- tional Publishing, 2021)

    Kayumov, Z., Tumakov, D. & Mosin, S.An Effect of Binarization on Handwritten Digits Recognition by Hierarchical Neural Networks, 94–106 (Springer Interna- tional Publishing, 2021). URL http://dx.doi.org/10.1007/978-3-030-84760-9 9

    work page doi:10.1007/978-3-030-84760-9 2021

  73. [73]

    S., Yang, G

    Huang, T. S., Yang, G. J. & Tang, G. Y. A fast two-dimensional median filtering algorithm.IEEE Transactions on Acoustics, Speech, and Signal Processing27, 13–18 (1981)

    1981

  74. [74]

    K.Digital Image Processing: PIKS Scientific Inside(Wiley, 2006)

    Pratt, W. K.Digital Image Processing: PIKS Scientific Inside(Wiley, 2006). URL http://dx.doi.org/10.1002/0470097434

    work page doi:10.1002/0470097434 2006

  75. [75]

    & Manduchi, R

    Tomasi, C. & Manduchi, R. Bilateral filtering for gray and color images (1998). 33 Fig. 5Evolution of classification accuracy and feature space properties with increasing occlusion size (percentage of target image area) for models trained using different scale augmentation ranges. (A-E) Trained with scale augmentationX, Y∼ U[10%,100%], (F-J) Trained with ...

    1998

  76. [76]

    8(A) - SNAN-CNN training protocol.(B) - SNAN-CNN testing protocol

    Learning rate 0.01 LR decay (γ c) Exponential decay 0.001 Epochs Training iterations 30 Augmentation Spatial translations±5 pixels T able 3Vision Transformer model parameters Component Description V alue/Size Architecture Parameters IInput image size (MNIST) 28×28×1 PPatch size 4×4 N t Number of patches 49 Dt Patch embedding dimension 128 Et Patch project...