Convolutional Spiking Neural Network for Image Classification

Andrey Lavrentyev; Mikhail Kiselev

REVIEW 4 major objections 4 minor 22 references

Convolutional Spiking Neural Network for Image Classification

T0 review · 4 major / 4 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 Fixed convolutional kernels give spiking net near-CNN accuracy

desk verdict Fixed convolutional kernels learned offline are a sensible idea, but the accuracy claim is inflated by learning kernels on the full dataset before the split. read the letter →

arxiv 2505.08514 v1 pith:FSIWI5YJ submitted 2025-05-13 cs.NE

Mikhail Kiselev , Andrey Lavrentyev This is my paper · ORCID

classification cs.NE

keywords convolutionalspikingneuralnetworkCoLaNETsynapticplasticityunsupervisedfeaturelearningimageclassificationNeovision2neuromorphichardwareweightsharing

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

This paper tries to establish that a spiking neural network can classify images nearly as well as a conventional convolutional network while leaving the convolutional weights completely fixed. Its strategy is to learn a small set of domain-level convolution kernels offline from unlabeled images, freeze them, and install them as non-plastic synaptic weights in a spiking convolutional front end. On the five-class Neovision2 benchmark the resulting network reaches 91.58% mean accuracy, against 92.34% for a similarly shaped trained CNN and 94.35% for a deeper CNN. If this holds, convolutional feature extraction becomes compatible with neuromorphic hardware constraints, because only the downstream classifier has to learn.

What carries the argument

The load-bearing mechanism is an offline, unsupervised kernel-learning algorithm applied to a collection G of unlabeled images from the target domain. It maintains a synaptic-resource tensor W for each of NC candidate K by K kernels; on each image it identifies the most strongly stimulated kernel positions, strengthens the resources of synapses connected to the bright pixels that caused firing, and subtracts an equal total from the other synapses so each neuron's resources are conserved. A winner-takes-all constraint prevents two kernels from learning the same image patch, forcing the matrices to become distinct detectors of frequent local structures such as edges, spots, and boundaries. After learning, the resource values are converted to weights by a bounded monotone mapping, scaled by a calibration constant, and fixed as the weights of rate-coded integrate-and-fire neurons in a convolutional layer. The classifier is a columnar layered spiking network (CoLaNET) that receives the flattened, mean-pooled feature maps and does all of the plastic learning.

What would settle it

Learn the 28 kernels on one subset of the Neovision2 clips, classify objects from a disjoint subset, and compare with kernels learned on the target subset; a large drop below the reported 91.58% would refute the transfer assumption, as would a marked accuracy gain if the same network is retrained with trainable convolution weights.

Watch

Extended reading notes

Core claim

The paper's central claim is that in-network learning of convolution kernels is unnecessary for a competitive spiking classifier. An unsupervised procedure, run on a representative set of unlabeled domain images, produces kernels that are ported as frozen weights into an SNN, and the full pipeline—frozen convolutional layer, average pooling, and a plastic columnar layered classifier—achieves 91.58% mean accuracy (SD 1.1) on Neovision2. This is within 0.76 percentage points of CNN2, the same-shape CNN trained end-to-end, and 2.77 points below the larger CNN1. The authors attribute the residual gap mainly to spike-rate encoding, which represents pixel brightness by only 11 spike counts rather than 256 gray levels, and point to the expected energy savings of neuromorphic execution as the compensating advantage.

Load-bearing premise

Convolution kernels learned from a representative unlabeled image set in one domain transfer to arbitrary classification tasks in that domain; the paper itself says this cannot be formally proven and is accepted from common sense.

Editorial extensions

If this is right

A spiking convolutional network can be built without weight sharing or in-network kernel training, removing a major obstacle to mapping convolutional feature extraction onto neuromorphic processors.
The same fixed kernel set can in principle be reused across classification tasks within one domain, with only the classifier retrained.
On Neovision2 the fixed-kernel SNN is competitive with a same-shape CNN, and it uses far fewer neurons (9,412 vs. 42,885 for the larger CNN), so the trade-off is a few points of accuracy against large hardware savings.
Because only the classifier layer is plastic, the network is suited to one-pass online learning without storing the full training set.

Reading between the lines

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

If the domain-transfer assumption holds, the same 28 kernels should work for other tasks built from the same camera stream; a direct test would be to learn kernels on one subset of Neovision2 clips and classify objects from another subset.
The kernel-learning algorithm is a batch analogue of STDP-based unsupervised feature learning, so applying it recursively to pooled feature maps could plausibly build deeper convolutional SNNs without violating locality.
The reported comparison uses one epoch and a small five-class task; whether the accuracy gap widens on larger, harder benchmarks such as full ImageNet subsets is unresolved by this paper.
Increasing the spike-count encoding from 11 levels to finer rate or temporal codes should narrow the gap to CNN1; this is testable by lengthening the image presentation window.

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.

Desk Editor's Note

read the letter

Quick take: the paper describes a way to build a convolutional SNN without weight sharing — kernels are learned offline on a large image set from the domain, then ported as fixed synapses into an SNN whose classifier is CoLaNET. The idea is genuinely useful for neuromorphic hardware where weight sharing is expensive. The kernel-learning algorithm itself is a reasonable standalone reworking of the STDP-based approach from [10], and the pipeline with CoLaNET is new.

Where it holds up: the algorithm is clearly specified, the kernel visualizations show structured features, and the experimental setup is straightforward. The authors are honest that the energy advantage is not measured and that the domain-transfer assumption is informal. For an engineering preprint, that's fair.

The soft spot is the evaluation. The kernels are learned on the entire 155,881-image set G, which is built from all Neovision2 clips, and then 5-fold cross-validation is run on the same images. So each test fold has already contributed to the fixed feature extractor. The CNN baselines are trained only on the training folds. That makes the 91.58% vs 92.34% comparison not apples-to-apples, and the 'comparable with traditional CNNs' claim is likely optimistic. This isn't a footnote issue; it changes the headline result. The fix is easy: re-learn kernels per training fold and report the mean and spread, or at least show that kernels learned on one clip set transfer to a disjoint clip set.

Other concerns are minor. Hyperparameters and the calibration coefficient are chosen heuristically on the same data; some ablation would help. No code is released, though a supplementary link exists. The energy claim is explicitly deferred, which is fine, but it means the main motivation remains unverified.

Who's it for? People building SNN pipelines for neuromorphic processors, and researchers working on locality-friendly learning rules. It deserves a serious referee, but I would not accept the accuracy numbers as they stand. I'd ask for the split-fair evaluation and ideally an ablation on kernel transferability before publication.

Referee Report

4 major / 4 minor

Summary. The paper presents a convolutional spiking neural network (CSNN) for image classification in which the convolutional kernels are fixed and learned offline from a set of unlabeled images from the target domain, thereby avoiding weight sharing during SNN operation. The kernels are learned by an iterative unsupervised algorithm (Section II) and ported into a single convolutional layer followed by mean pooling; classification is performed by the CoLaNET spiking classifier. The method is evaluated on a 5-class object classification task derived from the Neovision2 benchmark, reporting mean accuracy 91.58% (SD 1.1) over 5-fold cross-validation, compared to 92.34% (SD 0.42) for a similarly shaped CNN (CNN2) and 94.35% (SD 0.32) for a larger CNN (CNN1). The authors conclude that the CSNN achieves accuracy comparable to traditional CNNs while preserving plasticity locality.

Significance. If the result holds, the paper offers a practical way to combine convolutional feature extraction with spiking networks on neuromorphic hardware, addressing the locality-principle objection to weight sharing. The kernel-learning algorithm is presented as a standalone procedure, and the evaluation includes two conventional CNN baselines. However, the strength of the evidence is limited by the evaluation protocol and by the untested domain-generality assumption. The paper would be a useful contribution to the neuromorphic SNN literature if these issues are resolved; the current version's central claim rests on a single dataset with a potentially optimistic accuracy estimate.

major comments (4)

[Section IV, paragraph 'In such a way, we created the set G...'] The set G used to learn the fixed convolution kernels is constructed from all 155,881 Neovision2 images before the 5-fold split is performed. Consequently, every test fold participates in building the feature extractor, so the reported 91.58% accuracy is measured under transductive leakage. The CNN baselines, by contrast, are trained only on training folds (one epoch), so the comparison in Table 1 is not apples-to-apples. The authors should either re-learn the kernels on each training fold only, or explicitly quantify the effect of this leakage (e.g., by comparing with kernels learned on a disjoint image set). This is load-bearing because the conclusion that CSNN is 'comparable' to CNN2 depends on the accuracy estimate.
[Section IV, hyperparameter choices (K, NC, s, B, l, wmin, wmax) and the proportionality coefficient] All algorithmic hyperparameters and the pooling-layer calibration coefficient are selected using the same Neovision2 data on which accuracy is evaluated. For example, B is set to the mean brightness of G, l is derived from NE and the assumed brightness scale, and wmin/wmax are based on the observed brightness distribution; the calibration coefficient is chosen to make the mean pooling firing frequency approximately 50 Hz, again on this dataset. No sensitivity analysis or independent validation is provided, and the CoLaNET hyperparameters (microcolumn count 22, learning rate 0.0035, weight bounds) are reported without a selection procedure. The reported accuracy may therefore be optimistically biased. The authors should describe how these values were chosen and provide evidence that the result is robust to reasonable variations.
[Section I, 'Domain-level convolutions' assumption] The central assumption that kernels learned from a representative image set G transfer to other classification tasks in the same domain is stated as 'can hardly be proven formally but it follows from common sense' (Section I). The paper provides no experiment that isolates transferability: kernels are learned from and evaluated on the same Neovision2 image collection. Because the entire fixed-feature pipeline collapses if kernels are task-specific rather than domain-generic, the authors should test this assumption, for example by learning kernels on one clip set and classifying objects from another clip set, or by varying the classification task while keeping the kernel-learning set fixed.
[Table 1 and Section V] The claim that CSNN accuracy is 'comparable' to CNN2 is based on a single 5-fold cross-validation run. The mean difference is 0.76 percentage points, but the standard deviations (1.1 for CSNN, 0.42 for CNN2) overlap, and no statistical test or repeated-run variability is reported. With only five folds, the difference may not be significant. The authors should provide a paired comparison across folds or multiple independent runs, and state whether the difference is statistically reliable.

minor comments (4)

[Figure captions] The captions for Figures 2 and 3 appear to be swapped: Figure 2 (color-coded matrices) is labeled 'Color coded values of the convolution matrices' and Figure 3 (the network diagram) is also labeled 'Color coded values of the convolution matrices'. The captions should be corrected.
[Section II, pseudo-code] The pseudo-code contains an unclear line 'cbrt ← 0 for all <b, r, t> for which r = p & t = q'; this appears to be a leftover variable that is never used. The authors should clarify or remove it, and also define the shape of the convolution output tensor c explicitly.
[Section IV, dataset composition] The paper does not report the class distribution of the 155,881 images or the number of examples per target class in the 5-class task. Class imbalance could affect both the CNN and CSNN results, and reporting it would help interpret the accuracy numbers.
[References] Reference [10] is cited as the basis of the kernel-learning algorithm, but the relation to that algorithm is only qualitative; a precise description of the differences (e.g., the threshold-based update rule and the WTA competition) would help reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: fixed convolutions are produced by a standalone unsupervised algorithm and the classifier is trained separately; the reported accuracy is measured, not constructed from its inputs.

full rationale

The paper's derivation chain is linear and self-contained: Section II defines an iterative unsupervised algorithm that maps an unlabeled image set G to convolution kernels w via the plasticity rule in Eq. (1); Section III ports those kernels as fixed synaptic weights; Section IV trains the CoLaNET classifier on labeled examples and measures accuracy. At no point is the target accuracy defined in terms of the kernels, nor are the kernels defined in terms of the class labels or the accuracy. The convolution-learning algorithm is not a renamed version of the classification result, and the comparison with CNN1/CNN2 is an external benchmark. Self-citations to CoLaNET [20] and ArNI-X [23] are to an off-the-shelf classifier and an emulator, not to the novelty claim, so they are not load-bearing in a circular way. The paper's own admission that the domain-level convolution assumption 'can hardly be proven formally but it follows from common sense' is an explicit statement of an assumption, not a hidden circular step. A legitimate methodological caveat exists: the kernels are learned from the full shuffled G before the 5-fold split, so test-fold images contribute to the fixed feature extractor, which can bias the 91.58% accuracy estimate. That is data leakage or evaluation bias, however, not a definitional or equation-level circularity, because the kernels are unsupervised and the supervised classifier is fit separately. On the circularity axis, the derivation does not reduce to its own inputs.

Assumptions & free parameters 10 free parameters · 5 assumptions · 0 invented entities

The central pipeline relies on a small set of hand-set hyperparameters, an unproven transferability assumption, and a hardware-efficiency premise; none are backed by independent evidence in this preprint.

free parameters (10)

Convolution size K = 9
Chosen to match planned 28 maps and input constraints; no search or sensitivity analysis reported.
Kernel count NC = 28
Selected so that 28x6x6 = 1008 features fit under CoLaNET's 1024-input limit.
Stride s = 2
Chosen with K to produce 12x12 maps before pooling.
Brightness threshold B = 26
Set to the mean pixel brightness of G; used in the kernel update rule.
Learning rate l = ~8e-6
Derived from heuristic l = 100/255/NE so single-pixel contributions can cause firing; no tuning or validation.
wmin and wmax = -0.006536 and 0.0196
Chosen by balancing the number of bright vs dark pixels and to make one pixel's contribution exceed threshold; the paper calls this choice 'least evident'.
CoLaNET microcolumns per column = 22
Reported as a hyperparameter without sensitivity analysis.
CoLaNET learning rate and weight limits = 0.0035, min -0.0628, max 0.152
Reported as fixed for the experiment; no tuning details.
Input-to-pooling calibration coefficient = not reported
Set so mean pooling firing frequency is near 50 Hz via a monotonic, almost linear calibration; exact value not given.
Image presentation and silence times = 10 ms and 10 ms
CoLaNET constraint; rate coding maps brightness to 0-10 spikes.

assumptions (5)

domain assumption Domain-level convolution transferability: convolution matrices should correspond to structural features common to images from a domain and can be reused across classification tasks.
Section I states this 'can hardly be proven formally but it follows from common sense.' If false, the fixed kernels will not transfer to new tasks.
domain assumption Weight sharing cannot be efficiently implemented on neuromorphic hardware, so convolutions must be non-plastic and predetermined.
Introduction and Conclusion assert this hardware claim about Loihi and AltAI; no measurement is provided in this paper.
ad hoc to paper The offline kernel-learning algorithm of Section II converges to typical structural features of the image domain.
No convergence or correctness proof is given; the paper only offers an informal interpretation in terms of winners and synaptic resources.
domain assumption An IF neuron with subtractive reset and tau = 1 passes spike counts exactly, which makes average pooling exact.
Relies on equation (2) and the chosen neuron model; this is a modeling assumption rather than a measured property.
domain assumption A mean input rate of about 1 spike per input per image is good for CoLaNET learning.
Stated in Section IV as the basis for calibration; no ablation tests alternative rates.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Convolutional Spiking Neural Network for Image Classification." pith.science (2026). https://pith.science/paper/FSIWI5YJ

@misc{pith2026250508514,
  author       = {Pith},
  title        = {Pith review of: Convolutional Spiking Neural Network for Image Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FSIWI5YJ}},
  note         = {Machine review of arXiv:2505.08514}
}

read the original abstract

We consider an implementation of convolutional architecture in a spiking neural network (SNN) used to classify images. As in the traditional neural network, the convolutional layers form informational "features" used as predictors in the SNN-based classifier with CoLaNET architecture. Since weight sharing contradicts the synaptic plasticity locality principle, the convolutional weights are fixed in our approach. We describe a methodology for their determination from a representative set of images from the same domain as the classified ones. We illustrate and test our approach on a classification task from the NEOVISION2 benchmark.

Figures

Figures reproduced from arXiv: 2505.08514 by the authors.

**Figure 1.** Creation of image set for obtaining the convolution matrices. a. An orginal image converted to monochrome with square object labels. b. The differential heat map. c. Examples of images used to obtain the convolution matrices (here, brightness is inverted). c a a b [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

**Figure 2.** Color coded values of the convolution matrices. a [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Color coded values of the convolution matrices. INPUT DECISION CLASS LABELS learning neurons WTA REWGATE BIASGATE OUT INPUT L L L WTA WTA WTA REWGATE REWGATE REWGATE OUT CLASS LABEL BIASGATE 10 20 10 1 From other class labels plastic connections connection types strong excitatory excitatory delayed excitatory strong inhibitory activating blocking dopamine DECISION [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: CoLaNET architecture. a. General view. b. Structure of one column. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 17 canonical work pages

[10]

STDP-based spiking deep convolutional neural networks for object recognition,

S.R. Kheradpisheh, M. Ganjtabesh, S.J. Thorpe, and T. Masquelier, “STDP-based spiking deep convolutional neural networks for object recognition,” Neural Networks, vol. 99, 2017, pp. 56-67

work page 2017
[1]

Energy efficiency and environmental impact of training LLMs ,

K. Prashant, M. Chandrika, “Energy efficiency and environmental impact of training LLMs ,” International Journal of Research Publication and Reviews, vol. 5(7), 2024, pp. 1424-1430

work page 2024
[2]

Deep spiking convolutional neural network trained with unsupervised spike-timing- dependent plasticity,

C. Lee, G. Srinivasan, P. Panda, and K. Roy, “Deep spiking convolutional neural network trained with unsupervised spike-timing- dependent plasticity,” IEEE Trans. Cognit. Dev. Syst. , vol. 11, 2019, pp. 384–394

work page 2019
[3]

Feature extraction using spiking convolutional neural networks ,

R. Vaila, J. Chiasson, and V. Saxena, “Feature extraction using spiking convolutional neural networks ,” Proc. Int. Conf. Neuromorph. Syst. 2019, https://doi.org/10.1145/3354265.3354279

arXiv 2019
[4]

Training deep spiking convolutional neural networks with STDP -based unsupervised pre - training followed by supervised fine -tuning,

C. Lee, P. Panda, G. Srinivasan, and K. Roy, “Training deep spiking convolutional neural networks with STDP -based unsupervised pre - training followed by supervised fine -tuning,” Front. Neurosci, 2018, 12:435

work page 2018
[5]

Deep spiking convolutional neural network for single object localization based on deep continuous local learning ,

S. Barchid, J. Mennesson, and C. Djeraba, “Deep spiking convolutional neural network for single object localization based on deep continuous local learning ,” Proc. International Conference on Content -Based Multimedia Indexing (CBMI) , 2021, https://doi.org/10.1109/cbmi50038.2021.9461880. b

arXiv 2021
[6]

Unsupervised speech recognition through spike-timing-dependent plasticity in a convolutional spiking neural network,

M. Dong, X. Huang, and B. Xu, “Unsupervised speech recognition through spike-timing-dependent plasticity in a convolutional spiking neural network,” PLOS ONE, 2018, 13(11): e0204596

work page 2018
[7]

A new spiking convolutional recurrent neural network (SCRNN) with applications to event-based hand gesture recognition ,

Y. Xing, G. Di Caterina, and J. Soraghan, “A new spiking convolutional recurrent neural network (SCRNN) with applications to event-based hand gesture recognition ,” Front. Neurosci , 2020, 14:590164

work page 2020

Show all 22 references

[8]

A reservoir-based convolutional spiking neural network for gesture recognition from DVS input,

A. George, D. Banerjee, S. Dey, A. Mukherjee, B. Purushothaman, “A reservoir-based convolutional spiking neural network for gesture recognition from DVS input,” Proc. IJCNN-2020, 2020, pp. 1-9

2020
[9]

Spiking convolutional neural networks for text classification,

Ch. Lv, J. Xu, and X. Zheng, “Spiking convolutional neural networks for text classification,” arXiv:2406.19230, 2024

2024 arXiv
[11]

A convolutional spiking neural network with adaptive coding for motor imagery classification,

X. Liao, Y. Wu, Z. Wang, D. Wang, and H. Zhang, “A convolutional spiking neural network with adaptive coding for motor imagery classification,” Neurocomputing, vol. 549, 2023, 126470

2023
[12]

SLAYER: spike layer error reassignment in time

S.B. Shrestha, G. Orchard, "SLAYER: spike layer error reassignment in time", arXiv:1810.08646, 2018

2018 arXiv
[13]

Convolutional spiking neural networks for spatio-temporal feature extraction,

A. Samadzadeh, et al. “Convolutional spiking neural networks for spatio-temporal feature extraction,” Neural Process Lett, vol. 55, 2023, pp. 6979–6995

2023
[14]

Loihi: A Neuromorphic Manycore Processor with On-Chip Learning

M. Davies et al. “Loihi: A Neuromorphic Manycore Processor with On-Chip Learning”, IEEE Micro, vol. 38, no. 1, 2018, pp. 82-99

2018
[15]

Neuromorphic processor Alt AI for energy- efficient computing,

N.V. Grishanov et al., “Neuromorphic processor Alt AI for energy- efficient computing,” Nanoindustry Russia, vol. 96, 2019, pp. 531-538

2019
[16]

Exploring parameter and hyper-parameter spaces of neuroscience models on high performance computers with learning to learn ,

A. Yegenoglu, et al. “Exploring parameter and hyper-parameter spaces of neuroscience models on high performance computers with learning to learn ,” Frontiers in computational neuroscience, vol. 16, 2022, 885207

2022
[17]

Rate coding vs. temporal coding - is optimum between?

M. Kiselev, “Rate coding vs. temporal coding - is optimum between?” Proc. IJCNN-2016, 2016, pp. 1355-1359

2016
[18]

A synaptic plasticity rule providing a unified approach to supervised and unsupervised learning,

M. Kiselev, “A synaptic plasticity rule providing a unified approach to supervised and unsupervised learning,” Proc. IJCNN-2017, 2017, pp. 3806-3813

2017
[19]

A preprocessing layer in spiking neural networks – structure, parameters, performance criteria ,

M. Kiselev, and A. Lavrentyev, “A preprocessing layer in spiking neural networks – structure, parameters, performance criteria ,” Proc. IJCNN-2019, 2019, paper N-19450

2019
[20]

CoLaNET -- a spiking neural network with columnar layered architecture for classification

M. Kiselev, “ CoLaNET -- a spiking neural network with columnar layered architecture for classification”, 2024, http://arxiv.org/abs/2409.01230

2024 arXiv
[21]

http://ilab.usc.edu/neo2/dataset/
[22]

Advancing neuromorphic computing with Loihi: a survey of results and outlook,

M. Davies et al., "Advancing neuromorphic computing with Loihi: a survey of results and outlook," Proceedings of the IEEE, vol. 109, no. 5, 2021, pp. 911-934

2021

Pith tools

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

[1] [10]

STDP-based spiking deep convolutional neural networks for object recognition,

S.R. Kheradpisheh, M. Ganjtabesh, S.J. Thorpe, and T. Masquelier, “STDP-based spiking deep convolutional neural networks for object recognition,” Neural Networks, vol. 99, 2017, pp. 56-67

work page 2017

[2] [1]

Energy efficiency and environmental impact of training LLMs ,

K. Prashant, M. Chandrika, “Energy efficiency and environmental impact of training LLMs ,” International Journal of Research Publication and Reviews, vol. 5(7), 2024, pp. 1424-1430

work page 2024

[3] [2]

Deep spiking convolutional neural network trained with unsupervised spike-timing- dependent plasticity,

C. Lee, G. Srinivasan, P. Panda, and K. Roy, “Deep spiking convolutional neural network trained with unsupervised spike-timing- dependent plasticity,” IEEE Trans. Cognit. Dev. Syst. , vol. 11, 2019, pp. 384–394

work page 2019

[4] [3]

Feature extraction using spiking convolutional neural networks ,

R. Vaila, J. Chiasson, and V. Saxena, “Feature extraction using spiking convolutional neural networks ,” Proc. Int. Conf. Neuromorph. Syst. 2019, https://doi.org/10.1145/3354265.3354279

arXiv 2019

[5] [4]

Training deep spiking convolutional neural networks with STDP -based unsupervised pre - training followed by supervised fine -tuning,

C. Lee, P. Panda, G. Srinivasan, and K. Roy, “Training deep spiking convolutional neural networks with STDP -based unsupervised pre - training followed by supervised fine -tuning,” Front. Neurosci, 2018, 12:435

work page 2018

[6] [5]

Deep spiking convolutional neural network for single object localization based on deep continuous local learning ,

S. Barchid, J. Mennesson, and C. Djeraba, “Deep spiking convolutional neural network for single object localization based on deep continuous local learning ,” Proc. International Conference on Content -Based Multimedia Indexing (CBMI) , 2021, https://doi.org/10.1109/cbmi50038.2021.9461880. b

arXiv 2021

[7] [6]

Unsupervised speech recognition through spike-timing-dependent plasticity in a convolutional spiking neural network,

M. Dong, X. Huang, and B. Xu, “Unsupervised speech recognition through spike-timing-dependent plasticity in a convolutional spiking neural network,” PLOS ONE, 2018, 13(11): e0204596

work page 2018

[8] [7]

A new spiking convolutional recurrent neural network (SCRNN) with applications to event-based hand gesture recognition ,

Y. Xing, G. Di Caterina, and J. Soraghan, “A new spiking convolutional recurrent neural network (SCRNN) with applications to event-based hand gesture recognition ,” Front. Neurosci , 2020, 14:590164

work page 2020