Pith. sign in

REVIEW 3 major objections 5 minor 33 references

Continual Learning with Columnar Spiking Neural Networks

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 Columnar spiking network holds 92% accuracy across ten sequential tasks.

desk verdict A serious but uneven empirical study: CoLaNET avoids catastrophic forgetting on disjoint-task Permuted MNIST, but the headline numbers are tuned on the test tasks and the tables contradict each other. read the letter →

arxiv 2506.17169 v2 pith:6CUBZDXK submitted 2025-06-20 cs.NE cs.AI

Denis Larionov , Nikolay Bazenkov , Mikhail Kiselev This is my paper · ORCID
classification cs.NEcs.AI
keywords continuallearningcatastrophicforgettingspikingneuralnetworkcolumnarorganizationlocalanti-Hebbianplasticitydopamine-modulatedPermutedMNIST
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 argues that a spiking neural network organized into cortical-like columns can learn a sequence of tasks without catastrophic forgetting, provided the tasks do not share features. On ten sequentially learned pixel-permuted MNIST tasks, the authors report 92.31% average accuracy and only 4.35% degradation on the first task after learning the other nine, against a 49% drop for a standard one-layer ANN baseline. The same network fails to control forgetting when tasks share features: on MNIST followed by EMNIST letters, forgetting reaches 43%, and 32% in the reverse direction. The contribution is a demonstration that columnar separation of microcolumns, rather than the biological learning rule alone, is what buys continual learning, and that this works precisely when task boundaries coincide with feature boundaries.

What carries the argument

The central object is a microcolumn: a small group of leaky integrate-and-fire neurons inside a class-specific column that competes via winner-take-all for the right to encode a subclass. Learning is carried by two opposed local rules operating on the same plastic synapses — dopamine-modulated potentiation strengthens synapses that fired shortly before a reward spike, while anti-Hebbian depression weakens synapses that fired without reward — together with an adaptive firing threshold $u_{tr} = u_{const} + \alpha \sum_i w_i^+$ that raises a neuron's threshold as its positive weights grow, and optional synaptic renormalization controlled by a virtual-synapse count $n_s$. This machinery lets each microcolumn become a dedicated detector; whether it stays dedicated to one task or remains available for new tasks is governed by $\alpha$ and $n_s$.

What would settle it

Run CoLaNET on a sustained task sequence with measurable feature overlap between successive tasks, for example MNIST followed by EMNIST as the paper itself does, and compare first-task forgetting: the paper reports 4.35% on permuted tasks but 43% and 32% on the shared-feature pair, so a result near 4% on a shared-feature sequence would contradict the paper's stated scope.

Watch

Extended reading notes

Core claim

CoLaNET's microcolumns dedicate themselves to subclasses of a single task: within each class column, winner-take-all competition selects one learning neuron, dopamine-modulated plasticity potentiates the synapses that produced the correct spike, and anti-Hebbian plasticity suppresses synapses that fired without reward. The adaptive firing threshold and synaptic renormalization tune how specialized these microcolumns become. On Permuted MNIST, specialization is high enough that columns trained on earlier tasks stop responding to later permuted inputs, so old accuracies stay nearly frozen; with 45 microcolumns and threshold coefficient $\alpha = 0.01$, the network keeps learning each new task at roughly 92% accuracy while first-task accuracy falls only from 93.30% to 88.95%. On MNIST/EMNIST, microcolumns trained on digits also respond to similar-looking letters, so the same mechanism produces 43% and 32% forgetting. The paper's claim is that columnar SNNs offer a workable, biologically plausible route to continual learning in the domain-incremental setting where tasks occupy disjoint feature spaces, and that hyperparameters $\alpha$ and the virtual-synapse count $n_s$ control the stability-plasticity balance.

Load-bearing premise

The headline result stands on Permuted MNIST's random permutations making every task feature-disjoint; the moment tasks share visual features, as in the paper's own MNIST/EMNIST run, forgetting jumps to 43% and 32%.

Editorial extensions

If this is right

  • If the scope condition holds — tasks share no features — CoLaNET reduces first-task forgetting to about 4% over ten tasks while holding average accuracy near 92%, with no replay buffer, no gradient, and no task identities during inference.
  • The stability-plasticity trade-off is directly visible in two knobs: raising the adaptive-threshold coefficient $\alpha$ freezes old tasks but starves new ones; raising the virtual-synapse count $n_s$ diverts renormalization and trades accuracy for plasticity.
  • On the Permuted MNIST benchmark, CoLaNET's 92.39% average accuracy with 0.94% forgetting at 45 microcolumns beats the replay-based methods compared in the paper, including their joint-training reference of 89.05%.
  • For tasks with shared features, the mechanism does not solve continual learning by itself, and the paper points to a hierarchical architecture as the next step, applying columns only at deep layers where representations are task-specific.

Reading between the lines

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

  • Because the low-forgetting result depends on random pixel permutations destroying shared structure, a natural testable extension is to build task sequences with controlled degrees of feature overlap and measure where forgetting begins to climb; the paper's MNIST/EMNIST numbers suggest the transition is sharp, not gradual.
  • The same columnar logic might be applied to other local-learning spiking networks: if dedicated microcolumns are what isolate tasks, then any architecture that forms hard winner-take-all cell assemblies could inherit similar stability without the exact dopamine/anti-Hebbian combination.
  • The authors' proposed hierarchical fix is essentially a prediction that shared low-level features should be learned once in early layers, with columns reserved for the differing deep representations; this could be tested on a single network with a frozen shallow feature extractor and columnar top layers.
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. This paper proposes CoLaNET, a columnar spiking neural network with local learning rules, and evaluates it on continual learning benchmarks. The experiments cover an online streaming task-based domain-incremental setting with ten Permuted MNIST tasks and a two-task MNIST/EMNIST protocol, comparing against a one-layer ANN baseline and several existing continual learning methods. The paper reports that CoLaNET, with tuned hyperparameters, achieves about 92% average accuracy and low forgetting on Permuted MNIST, while acknowledging that this success depends on the tasks lacking shared feature structure; on MNIST/EMNIST, forgetting is much higher (43% and 32%). The source code is made available.

Significance. If the quantitative claims were reproducible, the paper would make a useful contribution to biologically plausible continual learning: it demonstrates a specific architectural mechanism (columnar separation plus adaptive threshold and synaptic renormalization) for trading off stability and plasticity without backpropagation or replay. The paper is honest about the restricted regime in which the method works, explicitly stating in Section 6 that Permuted MNIST results do not generalize to tasks with shared features. The open-source code and the detailed degradation profiles are strengths. However, the central quantitative claim is currently undercut by numerical inconsistencies between the aggregation table and the detailed profiles, and by in-sample hyperparameter selection, so the significance cannot be fully assessed until these issues are resolved.

major comments (3)
  1. [§5.1, Tables 3–5] The metrics in Table 3 are not consistent with the corresponding degradation profiles. For the 45-microcolumn configuration, the final row of Table 5 yields an average accuracy of 91.76% and a forgetting measure of 1.27, whereas Table 3 lists AA=92.31 and FM=1.19, and Table 7 reports AA=92.39±0.13 and FM=0.94±0.17. Moreover, the AIA values in Table 3 (e.g., 30.02 for the 45-microcolumn row) are implausibly low for the definition in Eq. (2), which should produce values in the 90s. Please clarify whether Table 3 aggregates multiple independent runs and, if so, provide the run-level data; otherwise correct the tables and the abstract, which cite these numbers.
  2. [§5.1, hyperparameter selection] The hyperparameter α is selected by sweeping values and inspecting the resulting AA and FM on the same ten Permuted MNIST tasks that are later reported (Figure 4; "We found the optimal α value ... to be 0.01"). No validation split or held-out task sequence is described. Consequently, the headline "92% accuracy" and "4% degradation" are the best point of an in-sample model-selection procedure, not an unbiased estimate for a new task sequence. The standard deviations in Table 7 are conditional on the selected α and do not capture the selection. Please add a validation procedure or report the full sensitivity of the results across a reasonable α range, and temper the abstract accordingly.
  3. [Abstract and §5.1] The statement "maintaining 92% accuracy on each" overstates the results. In Table 5, after training on all ten tasks, task 1 accuracy is 88.95% and task 10 is 91.22%, so not every task reaches 92%. The average accuracy across tasks is 91.76% in that run (or 92.31% as stated in Table 3). Please rephrase to "an average accuracy of about 92%" or give the exact minimum per-task accuracy, and clarify the "only 4% performance degradation" figure, which is 4.35 percentage points on task 1 in Table 5.
minor comments (5)
  1. [Table 3] The row label "45 microcolumns, α = 0.1" contradicts the text in §5.1, which states that the optimal α for the 45-microcolumn configuration is 0.01; please correct this typo.
  2. [References] Reference [12] lists the arXiv identifier as "arXiv preprint arXiv:12403.05175"; this appears to be a typo, likely "2403.05175".
  3. [Figure 1] The caption does not define the abbreviation WTA; please add a short explanation of the winner-take-all mechanism for readers unfamiliar with the CoLaNET architecture.
  4. [Tables 1–5] The column heading "Iterations" is used to denote training steps on successive tasks; consider renaming it to "Tasks" or "Training stage" for clarity.
  5. [Equation (2)] The reported AIA values in Table 3 are not computed according to Eq. (2), which defines AIA as the average of AA over task indices; please verify that the metric values in Table 3 are correct and consistent with the definitions.

Circularity Check

1 steps flagged · score 6.0 of 10

Permuted MNIST headline is a selected, not predicted, outcome: α is tuned on the same ten test tasks, so the reported 92%/4% figures are the best point of the sweep rather than an unbiased estimate.

  1. fitted input called prediction [Section 5.1, stage two/three paragraphs (near 'We found the optimal α value ... to be 0.01'); Figure 4; Tables 3, 5]
    "In the third stage, we repeated the Permuted MNIST experiment for selected α values with triple the number of microcolumns (from 15 to 45 microcolumns per column). We found the optimal α value for the 45-microcolumn configuration to be 0.01, which maintains learning plasticity across all 10 tasks (Table 3). Table 5 shows the degradation profile of CoLaNET withα = 0.01 and 45 microcolumns."

    Table 3/Table 5 report the ten-task result for the α and microcolumn count chosen by inspecting AA/FM on those same ten tasks: stage two sweeps α and plots AA/FM (Figure 4), and stage three 'found the optimal α value ... to be 0.01' from those curves. No held-out permutation set or validation split is described, so the headline 92.31% AA / 1.19% FM is the best point of a model-selection sweep over the test data, not an unbiased prediction. The Table 7 standard deviations are at a fixed, already-selected α and do not cover selection across configurations. The reported 'maintaining 92% accuracy' / '4% degradation' is thus the selected optimum, not an independent estimate.

full rationale

The central derivation is not self-definitional: CoLaNET is an existing, previously published architecture (refs. [7]-[9], [26]) with open-source code, and the paper honestly reports the failure mode on shared-feature tasks (Section 5.2, Section 6). The Permuted MNIST conclusion is explicitly conditional on task-disjointness, stated in the paper as 'Permuted MNIST simplifies stability-plasticity trade-off analysis in CoLaNET by eliminating inter-task interference through pattern independence.' That admission prevents a charge that the model secretly assumes its own result. The circularity that does exist is confined to the headline quantitative claim: the 92%/4% figures are selected by tuning α and microcolumn count on the same ten tasks used to evaluate them, making the central reported numbers a fitted input rather than an out-of-sample prediction. The self-citations to prior CoLaNET papers are legitimate prior work, not an unverified uniqueness theorem, so they do not add circularity by themselves. Score 6 reflects that the headline result reduces by construction to model selection on the test tasks; it does not reach 8-10 because the architecture content and the E/MNIST experiments remain independent and honestly reported.

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

The central results depend on tuned hyperparameters alpha and ns, and on the task-disjointness assumption. No new theoretical entities are introduced; CoLaNET is taken from the authors' prior work.

free parameters (3)
  • alpha (adaptive threshold coefficient) = 0.005 (15 microcolumns), 0.01 (45 microcolumns), 0.023817 (default from [8])
    Chosen by sweeping on Permuted MNIST to balance average accuracy and forgetting; the chosen value changes with microcolumn count, so it is a tuned hyperparameter, not derived.
  • ns (virtual synapse count) = 0 to 100k; best E/MNIST uses 100k
    Swept on the two-task E/MNIST benchmark; controls synaptic renormalization and directly affects forgetting.
  • microcolumns per column = 15 and 45
    Capacity hyperparameter; 45 selected for the ten-task experiment to improve plasticity. Not derived from theory.
assumptions (3)
  • domain assumption Random pixel permutation makes Permuted MNIST tasks feature-disjoint, so a columnar network can allocate separate microcolumns to each task without interference.
    This is the condition under which the headline 92% accuracy and 4% forgetting result is obtained. The paper itself notes the assumption does not hold for MNIST/EMNIST (Section 6).
  • domain assumption The local plasticity rules (anti-Hebbian depression, dopamine-modulated potentiation, WTA competition) are sufficient to learn the image classification tasks.
    The architecture's learning is assumed to converge to useful weights; the paper does not prove convergence, only shows empirical results.
  • standard math LIF neuron model with rate coding preserves enough information from pixel intensities for classification.
    Standard SNN modeling assumption used without proof here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Continual Learning with Columnar Spiking Neural Networks." pith.science (2026). https://pith.science/paper/6CUBZDXK

@misc{pith2026250617169,
  author       = {Pith},
  title        = {Pith review of: Continual Learning with Columnar Spiking Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6CUBZDXK}},
  note         = {Machine review of arXiv:2506.17169}
}
read the original abstract

Continual learning is a key feature of biological neural systems, but artificial neural networks often suffer from catastrophic forgetting. Instead of backpropagation, biologically plausible learning algorithms may enable stable continual learning. This study proposes columnar-organized spiking neural networks (SNNs) with local learning rules for continual learning and catastrophic forgetting. Using CoLaNET (Columnar Layered Network), we show that its microcolumns adapt most efficiently to new tasks when they lack shared structure with prior learning. We demonstrate how CoLaNET hyperparameters govern the trade-off between retaining old knowledge (stability) and acquiring new information (plasticity). We evaluate CoLaNET on two benchmarks: Permuted MNIST (ten sequential pixel-permuted tasks) and a two-task MNIST/EMNIST setup. Our model learns ten sequential tasks effectively, maintaining 92% accuracy on each. It shows low forgetting, with only 4% performance degradation on the first task after training on nine subsequent tasks.

Figures

Figures reproduced from arXiv: 2506.17169 by the authors.

Figure 1
Figure 1. CoLaNET architecture, inference regime. The image is presented over 10 time steps, followed by 10 time [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. CoLaNET architecture, learning regime. Learning is governed by three factors: anti-Hebbian plasticity (weak [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Two random pixel permutations of digit 5. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Average Accuracy and Forgetting Measure with different [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Receptive fields of plastic neurons after training on MNIST and subsequent training on EMNIST. Blue [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 22 canonical work pages

  1. [1]

    Mark B. Ring. Child: A first step towards continual learning. Machine Learning, 28(1):77–104, 1997

    work page 1997

  2. [2]

    Dhireesha Kudithipudi and et. al. Biological underpinnings of lifelong learning machines. Nature Machine Intelligence, 4, 2022

    work page 2022

  3. [3]

    Michael McCloskey and Neal J. Cohen. Catastrophic interference in connectionist networks: The sequential learning problem. Psychology of learning and motivation, 24:109–165, 1989

    work page 1989

  4. [4]

    Connectionist models of recognition memory: Constraints imposed by learning and forgetting functions

    Roger Ratcliff. Connectionist models of recognition memory: Constraints imposed by learning and forgetting functions. Psychological Review, 97(2):285–308, 1990

    work page 1990

  5. [5]

    Spike Neural Networks: Information Representation, Learning, Memory

    Mikhail Kiselev. Spike Neural Networks: Information Representation, Learning, Memory. Palmarium Academic Publishing, 2020

    work page 2020

  6. [6]

    Antonov, K.V

    D.I. Antonov, K.V . Sviatov, and S. Sukhov. Continuous learning of spiking networks trained with local rules. Neural Networks, 155:512–522, 2022. 12 A PREPRINT - S EPTEMBER 10, 2025

    work page 2022

  7. [7]

    Colanet – a spiking neural network with columnar layered architecture for classification

    Mikhail Kiselev. Colanet – a spiking neural network with columnar layered architecture for classification. arXiv preprint arXiv:2409.01230, 2024

    arXiv 2024

  8. [8]

    Classifying images with colanet spiking neural network - the mnist example

    Mikhail Kiselev. Classifying images with colanet spiking neural network - the mnist example. arXiv preprint arXiv:2409.07833, 2024

    arXiv 2024

Show all 33 references
  1. [9]

    A digital machine learning algorithm simulating spiking neural network CoLaNET

    Mikhail Kiselev. A digital machine learning algorithm simulating spiking neural network CoLaNET. arXiv preprint arXiv:2503.17111, 2025

    2025 arXiv

  2. [10]

    Goodfellow, Mehdi Mirza, Da Xiao, Aaron Courville, and Yoshua Bengio

    Ian J. Goodfellow, Mehdi Mirza, Da Xiao, Aaron Courville, and Yoshua Bengio. An empirical investigation of catastrophic forgetting in gradient-based neural networks. arXiv preprint arXiv:1312.6211, 2013

    2013 arXiv

  3. [11]

    Emnist: an extension of mnist to handwritten letters

    Gregory Cohen, Saeed Afshar, Jonathan Tapson, and André van Schaik. Emnist: an extension of mnist to handwritten letters. arXiv preprint arXiv:1702.05373, 2017

    2017 arXiv

  4. [12]

    van de Ven, Nicholas Soures, and Dhireesha Kudithipudi

    Gido M. van de Ven, Nicholas Soures, and Dhireesha Kudithipudi. Continual learning and catastrophic forgetting. arXiv preprint arXiv:12403.05175, 2024

    2024 arXiv

  5. [13]

    A comprehensive survey of continual learning: Theory, method and application

    Liyuan Wang, Xingxing Zhang, Hang Su, and Jun Zhu. A comprehensive survey of continual learning: Theory, method and application. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(8):5362–5383, 2024

    2024

  6. [14]

    A continual learning survey: Defying forgetting in classification tasks

    Matthias De Lange, Rahaf Aljundi, Marc Masana, Sarah Parisot, Xu Jia, Aleš Leonardis, Gregory Slabaugh, and Tinne Tuytelaars. A continual learning survey: Defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(7):3366–3385, 2022

    2022

  7. [15]

    Costajussa

    Magdalena Biesialska, Katarzyna Biesialska, and Marta R. Costajussa. Continual lifelong learning in natural language processing: A survey. In Proceedings of the 28th International Conference on Computational Linguistics, pages 6523–6541, 2020

    2020

  8. [16]

    Continual learning for large language models: A survey

    Tongtong Wu, Linhao Luo, Yuan-Fang Li, Shirui Pan, Thuy-Trang Vu, and Gholamreza Haffari. Continual learning for large language models: A survey. ArXiv, abs/2402.01364, 2024

    2024 arXiv

  9. [17]

    Towards continual reinforcement learning: A review and perspectives

    Khimya Khetarpal, Matthew Riemer, Irina Rish, and Doina Precup. Towards continual reinforcement learning: A review and perspectives. Journal of Artificial Intelligence Research, 75:1401–1476, 2022

    2022

  10. [18]

    Arslan Chaudhry, Marcus Rohrbach, Mohamed Elhoseiny, Thalaiyasingam Ajanthan, Puneet Kumar Dokania, Philip H. S. Torr, and Marc’Aurelio Ranzato. On tiny episodic memories in continual learning. arXiv: Learning, 2019

    2019

  11. [19]

    Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A

    James Kirkpatrick, Razvan Pascanu, Neil C. Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A. Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, Demis Hassabis, Claudia Clopath, Dharshan Kumaran, and Raia Hadsell. Overcoming catastrophic forgettin...

    2016

  12. [20]

    Continual learning through synaptic intelligence

    Friedemann Zenke, Ben Poole, and Surya Ganguli. Continual learning through synaptic intelligence. Proceedings of machine learning research, 70, 2017

    2017

  13. [21]

    Understanding the role of training regimes in continual learning

    Seyed Iman Mirzadeh, Mehrdad Farajtabar, Razvan Pascanu, and Hassan Ghasemzadeh. Understanding the role of training regimes in continual learning. ArXiv, 2020

    2020

  14. [22]

    Efficient lifelong learning with a-gem

    Arslan Chaudhry, Marc’Aurelio Ranzato, Marcus Rohrbach, and Mohamed Elhoseiny. Efficient lifelong learning with a-gem. ArXiv, 2018

    2018

  15. [23]

    Processing of expected and unexpected events during conditioning and attention: a psy- chophysiological theory

    Stephen Grossberg. Processing of expected and unexpected events during conditioning and attention: a psy- chophysiological theory. Psychological Review, 89(5):529–572, 1982

    1982

  16. [24]

    Mitigating forgetting in online continual learning with neuron calibration

    Haiyan Yin and Peng Yang. Mitigating forgetting in online continual learning with neuron calibration. In Neural Information Processing Systems, 2021

    2021

  17. [25]

    van de Ven and Andreas S

    Gido M. van de Ven and Andreas S. Tolias. Three scenarios for continual learning. arXiv preprint arXiv:1904.07734, 2019

    1904 arXiv

  18. [26]

    ArNI-X: A simple but powerful and flexible simulator for spiking neural networks, 2020

    Mikhail Kiselev. ArNI-X: A simple but powerful and flexible simulator for spiking neural networks, 2020

    2020

  19. [27]

    Gradient episodic memory for continual learning

    David Lopez-Paz and Marc’Aurelio Ranzato. Gradient episodic memory for continual learning. In Neural Information Processing Systems, 2017

    2017

  20. [28]

    Continual learning on permuted MNIST, 2024

    Denis Larionov. Continual learning on permuted MNIST, 2024

    2024

  21. [29]

    Continual learning on EMNST, 2025

    Denis Larionov. Continual learning on EMNST, 2025

    2025

  22. [30]

    Learning to learn without forgetting by maximizing transfer and minimizing interference

    Matthew Riemer, Ignacio Cases, Robert Ajemian, Miao Liu, Irina Rish, Yuhai Tu, and Gerald Tesauro. Learning to learn without forgetting by maximizing transfer and minimizing interference. ArXiv, 2018. 13 A PREPRINT - S EPTEMBER 10, 2025

    2018

  23. [31]

    Online continual learning with maximally interfered retrieval

    Rahaf Aljundi, Lucas Caccia, Eugene Belilovsky, Massimo Caccia, Min Lin, Laurent Charlin, and Tinne Tuytelaars. Online continual learning with maximally interfered retrieval. ArXiv, 2019

    2019

  24. [32]

    Quang Hong Pham, Chenghao Liu, Doyen Sahoo, and Steven C. H. Hoi. Contextual transformation networks for online continual learning. In International Conference on Learning Representations, 2021

    2021

  25. [33]

    Spyketorch: Efficient simulation of convolutional spiking neural networks with at most one spike per neuron

    Milad Mozafari, Mohammad Ganjtabesh, Abbas Nowzari-Dalini, and Timothée Masquelier. Spyketorch: Efficient simulation of convolutional spiking neural networks with at most one spike per neuron. Frontiers in Neuroscience, 13, 2019. 14

    2019

Pith tools

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