Pith. sign in

REVIEW 3 major objections 4 minor 64 references

TS-SNN: Temporal Shift Module for Spiking Neural Networks

T0 review · 3 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 A zero-FLOP temporal shift lets spiking networks see past and future spikes in one timestep, lifting accuracy on CIFAR-10, CIFAR-100, ImageNet, and CIFAR10-DVS.

desk verdict A competent transfer of TSM to SNNs with good ablations, but the 'future features' claim assumes offline processing; the causal gap should be tested. read the letter →

arxiv 2505.04165 v5 pith:VEDL4ZN4 submitted 2025-05-07 cs.NE cs.AI

Kairong Yu , Tianqing Zhang , Qi Xu , Gang Pan , Hongwei Wang This is my paper · ORCID
classification cs.NEcs.AI
keywords spikingneuralnetworkstemporalshiftfeaturefusionenergy-efficientcomputingsurrogategradienttrainingevent-basedvisionimageclassificationresidualconnection
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 claims that a simple temporal shift of spike-feature channels can give spiking neural networks (SNNs) much of the temporal modeling power they otherwise lack, at nearly zero computational cost. The proposed TS module shifts part of the channels one timestep forward, part one timestep backward, and leaves the rest in place, so that each timestep's computation sees a blend of past, present, and future spikes. A scaled residual connection adds the shifted tensor back to the original features using one learnable parameter, which prevents information loss and stabilizes training. On CIFAR-10, CIFAR-100, ImageNet, and the event-based CIFAR10-DVS, the resulting TS-SNN reports state-of-the-art accuracies with fewer timesteps than prior SNN methods. The claim matters because it suggests SNNs can close part of the accuracy gap with ANNs without sacrificing their low-energy advantage.

What carries the argument

The load-bearing object is the Temporal Shift module defined by Equation (6): channels are split into $C_k$ groups of size $C_{\text{fold}} = C/C_k$, two random split indices $g_1 < g_2$ are drawn, and the tensor is reassembled as $Z^{+1}=X[t+1, :g_1 \cdot C_{\text{fold}}, :, :]$, $Z^{-1}=X[t-1, g_1 \cdot C_{\text{fold}} : g_2 \cdot C_{\text{fold}}, :, :]$, and $Z^{0}=X[t, g_2 \cdot C_{\text{fold}} :, :, :]$, with zeros padding the temporal boundaries. This rearrangement turns each timestep's input into a channel-wise mixture of past, present, and future spikes at zero FLOP cost. The second piece is the residual combination $Z' = \alpha \odot Z + X$, where $\alpha$ is the single learnable penalty factor that weighs the shifted features against the original ones, which prevents the original current-timestep information from being lost and keeps training stable.

What would settle it

Train the same TS-SNN on CIFAR-100 with a causal variant that replaces the $+1$ shift with a second $-1$ shift, using only past and present features, and compare accuracy at timesteps 2 and 4. If the causal variant matches the reported 80.28% accuracy, the future-access assumption is not load-bearing; if accuracy drops substantially, the reported gains depend on seeing future spikes at the current timestep.

Watch

Extended reading notes

Core claim

The paper's central claim is that fusing spike features across adjacent timesteps via a channel shift is enough to substantially improve SNN accuracy, and that this fusion can be done with only one additional learnable parameter and negligible extra compute. Concretely, for an input tensor $X$ of shape $T \times C \times H \times W$, the module splits channels into segments and forms $Z$ with $Z^{+1}=X[t+1,\ldots]$, $Z^{-1}=X[t-1,\ldots]$, and $Z^{0}=X[t,\ldots]$, padding zeros and truncating at the temporal edges, and then outputs $Z' = \alpha \odot Z + X$ with $\alpha$ a scalar penalty factor (typically 0.2–0.5). The authors show that this module, inserted into ResNet and Spikeformer backbones, reaches 96.72% on CIFAR-10, 80.28% on CIFAR-100, 70.61% on ImageNet, and 83.90% on CIFAR10-DVS, often at smaller timesteps than competing SNNs, while consuming about 5.857 mJ per ImageNet image in their 45-nm energy model compared with 13.10 mJ for a SEW ResNet-18 baseline.

Load-bearing premise

The module assumes that at timestep $t$ the network can already read spike features from timestep $t+1$, which is only true if the full spike sequence is buffered or the input is processed offline; in strictly causal online inference that future information does not exist yet.

Editorial extensions

If this is right

  • The TS module can be dropped into existing SNN backbones such as ResNet and Spikeformer with only one extra parameter, so any accuracy gain it provides is nearly free in terms of model size and compute.
  • Since the shift operation itself is a zero-FLOP rearrangement, TS-SNN preserves the accumulation-only energy advantage of SNNs over ANNs; the reported ImageNet energy is 5.857 mJ per image.
  • With timesteps of 1–4 on static images and 10–16 on event data, TS-SNN shows that strong SNN accuracy does not require long temporal sequences.
  • The authors report that consistent application of the TS module during both training and inference, rather than training only, improves accuracy without changing FLOPs; on CIFAR10-DVS it rises from 81.20% to 83.60%.
  • At $T=1$ the shift is a no-op, yet accuracy still improves slightly; the paper attributes this to the learnable scaling factor $\alpha$, which can perturb features even when no temporal shift occurs.

Reading between the lines

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

  • The future-shift term $Z^{+1}=X[t+1]$ means the reported accuracies describe an offline or buffered setting; a causal, online version of TS-SNN would need either a one-timestep output delay or a redesigned shift that uses only past and present features, and its accuracy may differ from the numbers reported.
  • Because the random split points are drawn per module, the effective receptive field in time varies across layers; this may act as a lightweight temporal dropout or augmentation, which could partly explain the gains beyond pure feature fusion.
  • A direct comparison between TS-SNN and an explicitly causal temporal fusion method, such as recurrent connections or causal convolutions over time, with matched compute would separate the benefit of temporal mixing per se from the benefit of seeing the future.
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 / 4 minor

Summary. The paper proposes TS-SNN, a temporal shift module for spiking neural networks. The module splits the channels of a spiking feature tensor into three groups, shifts one group forward in time, one backward, and leaves one unchanged, then combines the shifted tensor with the original input through a residual term scaled by a learnable scalar alpha. The authors report state-of-the-art accuracy on CIFAR-10 (96.72%), CIFAR-100 (80.28%), ImageNet (70.61%), and CIFAR10-DVS (83.90%) with fewer timesteps than prior SNN methods, and they provide energy estimates based on MAC/AC counts. The paper includes ablations on the channel folding factor, shift direction combinations, fixed versus random split strategies, and train/inference consistency, plus a small experiment inserting the module into a transformer-based SNN.

Significance. If the central claims hold, the TS module is an attractive plug-in: it adds a single scalar parameter per insertion, has essentially no arithmetic cost, and improves accuracy across static and event-based benchmarks while preserving the sparsity of SNN computation. The paper has concrete strengths: results are reported with standard deviations over three trials, the ablation set is reasonably broad, the transformer-based generality test is valuable, and the energy accounting follows a standard MAC/AC estimation framework used in the SNN literature. The main risk is not internal inconsistency of the shift algebra, but whether the method as defined is deployable in the online, low-latency settings implied by the paper's framing, and whether the reported gains are correctly attributed to the temporal shift rather than to training details or test-set-based hyperparameter selection.

major comments (3)
  1. [Eq. (6), Algorithm 1, Section 4.3] The future-branch of the shift, Z+1 = X[t+1, :g1*Cfold, :, :], requires spike features from timestep t+1 to be available when the module computes its output for timestep t. In a causal streaming deployment, which is the setting implied by 'low-latency' SNN inference and by the CIFAR10-DVS event-stream experiments, those features do not exist yet. The paper does not state whether all experiments were run in an offline full-sequence mode, whether one timestep of buffering/delay per shifted layer was introduced, or whether some other protocol was used. This is load-bearing because the claimed advantage over prior SNNs is explicitly tied to fusing future features within one timestep while maintaining low energy cost. Please specify the inference protocol precisely, quantify the resulting latency and memory overhead, and report results for a causal variant in which the +1 branch is replaced by a current-timestep or delayed branch.
  2. [Section 4.2, Tables 1 and 4] There is no no-TS baseline under the same training pipeline for the main ResNet-19/ResNet-20 experiments. The ablations vary Ck, shift directions, fixed versus random splits, and train/inference consistency, but none removes the TS module entirely in the ResNet setup; the only direct with/without-TS comparison is Table 4 for the transformer architecture. Because Table 1 compares against methods with different training recipes, and because the T=1 results already show that alpha alone contributes non-trivially, the reported improvements cannot be cleanly attributed to the temporal shift operation as opposed to the overall training setup or the residual alpha scaling. Please add a same-recipe no-TS baseline for ResNet-19/20 and, ideally, a no-TS baseline that retains the alpha scaling, to isolate the contribution of the shift itself.
  3. [Figure 3, Figure 5, Appendix A.3] The hyperparameters Ck and the shift direction combination appear to be selected using test-set accuracy. Figure 3 reports CIFAR-100 test accuracy as a function of Ck and selects Ck=32; Figure 5 selects the L-R-0 direction combination from test accuracy; Appendix A.3 describes dataset splits but no validation split for model selection. Selecting hyperparameters on the test set introduces optimistic bias into the reported SOTA numbers, which are the paper's central quantitative claim. Please use an independent validation split for hyperparameter selection and report the test results obtained with the selected configuration, or provide a nested evaluation that accounts for the selection procedure.
minor comments (4)
  1. [Algorithm 1 and Section 4.2] The procedure 'Generate random indices If for channel groups' is underspecified: it is unclear whether g1 and g2 are resampled on every forward pass, fixed after initialization, or chosen per training run. Since the random-versus-fixed ablation is claimed to matter (71.63% vs 71.24%), please state the sampling rule and the seed/reproducibility policy, and clarify whether inference is stochastic.
  2. [Section 3.2 and Section 4.5] The claim that the TS operation introduces 'no additional computational burden' counts only arithmetic operations. The shift still requires data movement and, in the offline full-sequence interpretation, buffering of activation tensors across timesteps. Please separate FLOP/energy arithmetic from memory and latency overhead, especially in Table 5 and in the energy-efficiency discussion.
  3. [Table 5] For ImageNet, Table 5 reports energy for 'SEW ResNet18' while Table 2 reports ImageNet accuracy for ResNet-18 and ResNet-34. Please clarify which architecture was used for the energy measurement, whether the TS module was inserted into that architecture, and how the comparison to the reported 13.10 mJ SEW ResNet18 baseline was made.
  4. [Appendix A.1 and Conclusion] There are several small presentation errors, including 'LIF Modal' in the appendix heading and 'an novel Temporal Shift Module' in the conclusion; these should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: TS-SNN is an empirical architecture study whose reported accuracy gains are measured, not derived from fitted constants, definitions, or load-bearing self-citations.

full rationale

The paper's central claims are empirical: the TS module is defined by Equations (5)-(7) and Algorithm 1, and its benefits are established through benchmark comparisons and ablations in Section 4. The claimed temporal fusion is indeed a direct consequence of the definition of the shift in Equation (6), but the paper does not derive any prediction from that definition; it reports measured accuracy. The residual combination Z' = alpha*Z + X is a fixed architectural choice, and alpha and Ck are hyperparameters tuned through ablations, not parameters fitted to test data and then renamed predictions. The T=1 improvement is attributed post hoc to alpha scaling, which is an explanation rather than a circular derivation. Several references are to the authors' prior work (e.g., Yu et al. 2025a,b; Zhang et al. 2025a,b), but these are context citations for SNN training and attention modules and are not load-bearing for the TS-SNN mechanism; no uniqueness theorem or ansatz is imported from those papers. The concern that Z+1 = X[t+1,...] requires future spikes is a real latency/causality limitation for online inference, but that is a correctness and deployment issue, not circularity: the reported numbers were obtained under the stated offline shift operation. No equation reduces to another by construction, and no fitted input is presented as a prediction. Therefore the circularity score is 0.

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

The central claim rests on standard SNN training assumptions, plus two paper-specific modeling choices: the availability of future timestep features and the random selection of channel split points. The free parameters alpha and Ck are tuned on benchmark accuracy, and the randomness of g1 and g2 is under-specified. No new physical entities are introduced.

free parameters (3)
  • alpha (penalty factor) = 0.5 for CIFAR-10/100, 0.2 for CIFAR10-DVS/ImageNet; learned during training
    Single learnable scalar in the residual combination Z' = alpha * Z + X (Equation 7). It controls how much shifted content is mixed with the original. At T=1 the shift is null, and the paper attributes the observed accuracy gain to alpha, making this parameter load-bearing for the claimed improvement.
  • Ck (channel folding factor) = 32
    Chosen via ablation on CIFAR-100 (Figure 3). Determines the size of channel groups that are shifted together. The paper does not report a validation split, so this choice appears to be tuned on the test benchmark.
  • g1, g2 (random split indices) = random per forward pass, not fitted
    Algorithm 1 generates random indices for channel group boundaries on each call. This stochastic design choice affects which channels are shifted left, right, or not at all, and no seed or distribution is specified.
assumptions (4)
  • standard math LIF neuron dynamics and surrogate gradient training are valid and sufficient for training the network.
    Equations (1)-(4) follow prior SNN training literature (STBP, surrogate gradient); the paper relies on these without re-derivation.
  • ad hoc to paper Future spike features X[t+1] are available at time t.
    Equation (6) uses X[t+1] at step t, assuming future inputs are accessible. This is not discussed and conflicts with online low-latency inference.
  • ad hoc to paper Random channel split indices are a valid and stable design choice.
    Algorithm 1 generates random indices per call without specifying distribution or seed, making the temporal shift stochastic and potentially non-reproducible.
  • standard math Energy estimates follow the 45nm, 0.9V model with 4.6 pJ per MAC and 0.9 pJ per AC.
    The energy analysis in Section A.4 relies on prior literature (Horowitz 2014, Qiao et al. 2015) without independent validation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TS-SNN: Temporal Shift Module for Spiking Neural Networks." pith.science (2026). https://pith.science/paper/VEDL4ZN4

@misc{pith2026250504165,
  author       = {Pith},
  title        = {Pith review of: TS-SNN: Temporal Shift Module for Spiking Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VEDL4ZN4}},
  note         = {Machine review of arXiv:2505.04165}
}
read the original abstract

Spiking Neural Networks (SNNs) are increasingly recognized for their biological plausibility and energy efficiency, positioning them as strong alternatives to Artificial Neural Networks (ANNs) in neuromorphic computing applications. SNNs inherently process temporal information by leveraging the precise timing of spikes, but balancing temporal feature utilization with low energy consumption remains a challenge. In this work, we introduce Temporal Shift module for Spiking Neural Networks (TS-SNN), which incorporates a novel Temporal Shift (TS) module to integrate past, present, and future spike features within a single timestep via a simple yet effective shift operation. A residual combination method prevents information loss by integrating shifted and original features. The TS module is lightweight, requiring only one additional learnable parameter, and can be seamlessly integrated into existing architectures with minimal additional computational cost. TS-SNN achieves state-of-the-art performance on benchmarks like CIFAR-10 (96.72\%), CIFAR-100 (80.28\%), and ImageNet (70.61\%) with fewer timesteps, while maintaining low energy consumption. This work marks a significant step forward in developing efficient and accurate SNN architectures.

Figures

Figures reproduced from arXiv: 2505.04165 by the authors.

Figure 1
Figure 1. Illustration of Temporal Shift. (a) Initial spike feature representation, where each row represents a different channel and each column represents a different timestep. (b) Temporal shift operation. Some channel are shifted forward (to the right) along the temporal dimension, while others are shifted backward (to the left), with the remaining features left unchanged. (c) To maintain consistent tensor dimensions afte… view at source ↗
Figure 2
Figure 2. Overview of the TS-SNN Architecture. unshifted, representing shifts of +1, -1, and 0 timesteps, respectively. The TS module is designed to efficiently model temporal dynamics within SNN by shifting portions of the feature map along the temporal dimension. The Temporal Shift operation splits the input feature tensor into multiple seg￾ments and then assigns two random split points to divide the original features into … view at source ↗
Figure 3
Figure 3. Impact of Channel Folding Factor (Ck) on CIFAR-100 Accuracy with ResNet20 architecture. The accuracy peaks when Ck is set to 32. are standard benchmarks for image classification, consist￾ing of 50,000 training images and 10,000 testing images, all sized 32 × 32. CIFAR-10 comprises 10 classes, while CIFAR-100 contains 100 classes. ImageNet (Deng et al., 2009) is a large-scale dataset extensively used for bench￾markin… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Comparison of Fixed-Length (left) and Scalable Random Transformations (right) in Temporal Feature Shift. Impact of Channel Folding Factor on Model Accuracy The principle of the TS module is to distort spike features along the temporal dimension. Different levels of dis…
Figure 6
Figure 6. Figure 6: Accuracy improvement with consistent TS Module ap￾plication across training and inference stages. figuration of left-right-no shift, yields the highest accuracy at 72.52%. Consequently, this combination (L-R-0) has been default shift direction combination in all experi…
Figure 7
Figure 7. Figure 7: Average Firing Rate of each layer in Vanilla SNN and TS SNN ImageNet. For the ImageNet dataset, the initial learning rate was set to 0.1, with a batch size of 64 and a total of 320 training epochs. The parameter Ck was set at 32, and α was initialized at 0.2. A.4. Anal…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 38 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    and Pereda, A

    Alcamí, P. and Pereda, A. E. Beyond plasticity: the dynamic impact of electrical synapses on neural circuits. Nature Reviews Neuroscience, 20 0 (5): 0 253--271, May 2019. ISSN 1471-0048. doi:10.1038/s41583-019-0133-5. URL https://www.nature.com/articles/s41583-019-0133-5

    work page doi:10.1038/s41583-019-0133-5 2019

  3. [3]

    and Poo, M.-m

    Bi, G.-q. and Poo, M.-m. Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type. Journal of neuroscience, 18 0 (24): 0 10464--10472, 1998. ISBN: 0270-6474 Publisher: Soc Neuroscience

    work page 1998

  4. [4]

    A Fully Spiking Hybrid Neural Network for Energy - Efficient Object Detection

    Chakraborty, B., She, X., and Mukhopadhyay, S. A Fully Spiking Hybrid Neural Network for Energy - Efficient Object Detection . IEEE Transactions on Image Processing, 30: 0 9014--9029, 2021. ISSN 1941-0042. doi:10.1109/TIP.2021.3122092. URL https://ieeexplore.ieee.org/document/9591302. Conference Name: IEEE Transactions on Image Processing

    work page arXiv 2021

  5. [5]

    Y., Lee, K.-Y., and Hsu, W

    Chang, Y.-L., Liu, Z. Y., Lee, K.-Y., and Hsu, W. Learnable Gated Temporal Shift Module for Deep Video Inpainting , July 2019. URL http://arxiv.org/abs/1907.01131. arXiv:1907.01131 [cs]

    arXiv 2019

  6. [6]

    Training Full Spike Neural Networks via Auxiliary Accumulation Pathway , January 2023

    Chen, G., Peng, P., Li, G., and Tian, Y. Training Full Spike Neural Networks via Auxiliary Accumulation Pathway , January 2023. URL http://arxiv.org/abs/2301.11929. arXiv:2301.11929 [cs]

    arXiv 2023

  7. [7]

    All You Need Is a Few Shifts : Designing Efficient Convolutional Neural Networks for Image Classification

    Chen, W., Xie, D., Zhang, Y., and Pu, S. All You Need Is a Few Shifts : Designing Efficient Convolutional Neural Networks for Image Classification . pp.\ 7241--7250, 2019. URL https://openaccess.thecvf.com/content_CVPR_2019/html/Chen_All_You_Need_Is_a_Few_Shifts_Designing_Efficient_Convolutional_CVPR_2019_paper.html

    work page 2019

  8. [8]

    Tensor Decomposition Based Attention Module for Spiking Neural Networks

    Deng, H., Zhu, R., Qiu, X., Duan, Y., Zhang, M., and Deng, L. Tensor Decomposition Based Attention Module for Spiking Neural Networks , October 2023. URL http://arxiv.org/abs/2310.14576. arXiv:2310.14576 [cs]

    work page Pith review arXiv 2023

Show all 64 references
  1. [9]

    ImageNet : A large-scale hierarchical image database

    Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. ImageNet : A large-scale hierarchical image database. In 2009 IEEE Conference on Computer Vision and Pattern Recognition , pp.\ 248--255, June 2009. doi:10.1109/CVPR.2009.5206848. URL https://ieeexplore.ieee.or...

    2009

  2. [10]

    Temporal Efficient Training of Spiking Neural Network via Gradient Re -weighting

    Deng, S., Li, Y., Zhang, S., and Gu, S. Temporal Efficient Training of Spiking Neural Network via Gradient Re -weighting. In Proceedings of ICLR , October 2021. URL https://openreview.net/forum?id=_XNtisL32jv

    2021

  3. [11]

    Dynamic Image Quantization Using Leaky Integrate -and- Fire Neurons

    Doutsi, E., Fillatre, L., Antonini, M., and Tsakalides, P. Dynamic Image Quantization Using Leaky Integrate -and- Fire Neurons . IEEE Transactions on Image Processing, 30: 0 4305--4315, 2021. ISSN 1941-0042. doi:10.1109/TIP.2021.3070193. URL https://ieeexplore.ieee.org/documen...

    2021

  4. [12]

    Temporal Effective Batch Normalization in Spiking Neural Networks

    Duan, C., Ding, J., Chen, S., Yu, Z., and Huang, T. Temporal Effective Batch Normalization in Spiking Neural Networks . Advances in Neural Information Processing Systems, 35: 0 34377--34390, December 2022. URL https://proceedings.neurips.cc/paper_files/paper/2022/hash/de2ad3ed...

    2022

  5. [13]

    Deep Residual Learning in Spiking Neural Networks

    Fang, W., Yu, Z., Chen, Y., Huang, T., Masquelier, T., and Tian, Y. Deep Residual Learning in Spiking Neural Networks . Proceedings of NeurIPS, 34: 0 21056--21069, 2021 a . URL https://proceedings.neurips.cc/paper/2021/hash/afe434653a898da20044041262b3ac74-Abstract.html

    2021

  6. [14]

    Incorporating Learnable Membrane Time Constant To Enhance Learning of Spiking Neural Networks

    Fang, W., Yu, Z., Chen, Y., Masquelier, T., Huang, T., and Tian, Y. Incorporating Learnable Membrane Time Constant To Enhance Learning of Spiking Neural Networks . In Proceedings of ICCV , pp.\ 2661--2671, 2021 b . URL https://openaccess.thecvf.com/content/ICCV2021/html/Fang_I...

    2021

  7. [15]

    and Zhao, J

    Guan, B. and Zhao, J. Relative pose estimation for multi-camera systems from point correspondences with scale ratio. In Proceedings of the 30th ACM International Conference on Multimedia, pp.\ 5036--5044, 2022

    2022

  8. [16]

    Minimal solutions for relative pose with a single affine correspondence

    Guan, B., Zhao, J., Li, Z., Sun, F., and Fraundorfer, F. Minimal solutions for relative pose with a single affine correspondence. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 1929--1938, 2020

    1929

  9. [17]

    Multi-dimensional pruning: A unified framework for model compression

    Guo, J., Ouyang, W., and Xu, D. Multi-dimensional pruning: A unified framework for model compression. In CVPR, 2020

    2020

  10. [18]

    Multidimensional pruning and its extension: A unified framework for model compression

    Guo, J., Xu, D., and Ouyang, W. Multidimensional pruning and its extension: A unified framework for model compression. IEEE Transactions on Neural Networks and Learning Systems, 2023 a

    2023

  11. [19]

    IM - Loss : Information Maximization Loss for Spiking Neural Networks

    Guo, Y., Chen, Y., Zhang, L., Liu, X., Wang, Y., Huang, X., and Ma, Z. IM - Loss : Information Maximization Loss for Spiking Neural Networks . Proceedings of NeurIPS, 35: 0 156--166, December 2022 a . URL https://proceedings.neurips.cc/paper_files/paper/2022/hash/010c5ba0cafc7...

    2022

  12. [20]

    RecDis - SNN : Rectifying Membrane Potential Distribution for Directly Training Spiking Neural Networks

    Guo, Y., Tong, X., Chen, Y., Zhang, L., Liu, X., Ma, Z., and Huang, X. RecDis - SNN : Rectifying Membrane Potential Distribution for Directly Training Spiking Neural Networks . In Proceedings of CVPR , pp.\ 326--335, New Orleans, LA, USA, June 2022 b . IEEE. ISBN 978-1-66546-9...

    2022

  13. [21]

    Real Spike : Learning Real - Valued Spikes for Spiking Neural Networks

    Guo, Y., Zhang, L., Chen, Y., Tong, X., Liu, X., Wang, Y., Huang, X., and Ma, Z. Real Spike : Learning Real - Valued Spikes for Spiking Neural Networks . In Proceedings of ECCV , Lecture Notes in Computer Science , pp.\ 52--68, Cham, 2022 c . Springer Nature Switzerland. ISBN ...

    2022 doi

  14. [22]

    RMP - Loss : Regularizing Membrane Potential Distribution for Spiking Neural Networks , August 2023 b

    Guo, Y., Liu, X., Chen, Y., Zhang, L., Peng, W., Zhang, Y., Huang, X., and Ma, Z. RMP - Loss : Regularizing Membrane Potential Distribution for Spiking Neural Networks , August 2023 b . URL http://arxiv.org/abs/2308.06787. arXiv:2308.06787 [cs]

    2023 arXiv

  15. [23]

    Membrane Potential Batch Normalization for Spiking Neural Networks

    Guo, Y., Zhang, Y., Chen, Y., Peng, W., Liu, X., Zhang, L., Huang, X., and Ma, Z. Membrane Potential Batch Normalization for Spiking Neural Networks . In Proceedings of the IEEE / CVF International Conference on Computer Vision ( ICCV ) , pp.\ 19420--19430, October 2023 c

    2023

  16. [24]

    and Roy, K

    Han, B. and Roy, K. Deep Spiking Neural Network : Energy Efficiency Through Time Based Coding . In Proceedings of ECCV , Lecture Notes in Computer Science , pp.\ 388--404, 2020. doi:10.1007/978-3-030-58607-2_23

    2020 doi

  17. [25]

    Deep Residual Learning for Image Recognition

    He, K., Zhang, X., Ren, S., and Sun, J. Deep Residual Learning for Image Recognition . pp.\ 770--778, 2016. URL https://openaccess.thecvf.com/content_cvpr_2016/html/He_Deep_Residual_Learning_CVPR_2016_paper.html

    2016

  18. [26]

    Hebb, D. O. The organization of behavior: A neuropsychological theory . Psychology press, 2005. ISBN 1-135-63191-3

    2005

  19. [27]

    1.1 Computing 's energy problem (and what we can do about it)

    Horowitz, M. 1.1 Computing 's energy problem (and what we can do about it). In 2014 IEEE International Solid - State Circuits Conference Digest of Technical Papers ( ISSCC ) , pp.\ 10--14, February 2014. doi:10.1109/ISSCC.2014.6757323. URL https://ieeexplore.ieee.org/abstract/...

    2014

  20. [28]

    Advancing Spiking Neural Networks Toward Deep Residual Learning

    Hu, Y., Deng, L., Wu, Y., Yao, M., and Li, G. Advancing Spiking Neural Networks Toward Deep Residual Learning . IEEE Transactions on Neural Networks and Learning Systems, pp.\ 1--15, 2024. ISSN 2162-2388. doi:10.1109/TNNLS.2024.3355393. URL https://ieeexplore.ieee.org/abstract...

    2024

  21. [29]

    and Kim, J

    Jeon, Y. and Kim, J. Constructing Fast Network through Deconstruction of Convolution . In Advances in Neural Information Processing Systems , volume 31. Curran Associates, Inc., 2018. URL https://proceedings.neurips.cc/paper/2018/hash/9719a00ed0c5709d80dfef33795dcef3-Abstract.html

    2018

  22. [30]

    Cifar-10 (canadian institute for advanced research)

    Krizhevsky, A., Nair, V., and Hinton, G. Cifar-10 (canadian institute for advanced research). URL http://www. cs. toronto. edu/kriz/cifar. html, 5 0 (4): 0 1, 2010

    2010

  23. [31]

    S., Panda, P., Srinivasan, G., and Roy, K

    Lee, C., Sarwar, S. S., Panda, P., Srinivasan, G., and Roy, K. Enabling Spike - Based Backpropagation for Training Deep Neural Network Architectures . Frontiers in Neuroscience, 14, 2020. ISSN 1662-453X. URL https://www.frontiersin.org/articles/10.3389/fnins.2020.00119

    2020

  24. [32]

    C., See, S., Wang, X., Qin, H., and Li, H

    Li, D., Shi, X., Zhang, Y., Cheung, K. C., See, S., Wang, X., Qin, H., and Li, H. A Simple Baseline for Video Restoration with Grouped Spatial - Temporal Shift . In 2023 IEEE / CVF Conference on Computer Vision and Pattern Recognition ( CVPR ) , pp.\ 9822--9832, Vancouver, BC,...

    2023

  25. [33]

    Cifar10-dvs: an event-stream dataset for object classification

    Li, H., Liu, H., Ji, X., Li, G., and Shi, L. Cifar10-dvs: an event-stream dataset for object classification. Frontiers in neuroscience, 11: 0 309, 2017 a

    2017

  26. [34]

    CIFAR10 - DVS : An Event - Stream Dataset for Object Classification

    Li, H., Liu, H., Ji, X., Li, G., and Shi, L. CIFAR10 - DVS : An Event - Stream Dataset for Object Classification . Frontiers in Neuroscience, 11, 2017 b . ISSN 1662-453X. URL https://www.frontiersin.org/articles/10.3389/fnins.2017.00309

    2017

  27. [35]

    Spikeformer: A Novel Architecture for Training High - Performance Low - Latency Spiking Neural Network , November 2022

    Li, Y., Lei, Y., and Yang, X. Spikeformer: A Novel Architecture for Training High - Performance Low - Latency Spiking Neural Network , November 2022. URL http://arxiv.org/abs/2211.10686. arXiv:2211.10686 [cs]

    2022 arXiv

  28. [36]

    Learnable Surrogate Gradient for Direct Training Spiking Neural Networks

    Lian, S., Shen, J., Liu, Q., Wang, Z., Yan, R., and Tang, H. Learnable Surrogate Gradient for Direct Training Spiking Neural Networks . In Proceedings of IJCAI , pp.\ 3002--3010, Macau, SAR China, August 2023. ISBN 978-1-956792-03-4. doi:10.24963/ijcai.2023/335. URL https://ww...

    2023 doi

  29. [37]

    IM - LIF : Improved Neuronal Dynamics With Attention Mechanism for Direct Training Deep Spiking Neural Network

    Lian, S., Shen, J., Wang, Z., and Tang, H. IM - LIF : Improved Neuronal Dynamics With Attention Mechanism for Direct Training Deep Spiking Neural Network . IEEE Transactions on Emerging Topics in Computational Intelligence, pp.\ 1--11, 2024. ISSN 2471-285X. doi:10.1109/TETCI.2...

    2024

  30. [38]

    TSM : Temporal Shift Module for Efficient Video Understanding

    Lin, J., Gan, C., and Han, S. TSM : Temporal Shift Module for Efficient Video Understanding . pp.\ 7083--7093, 2019. URL https://openaccess.thecvf.com/content_ICCV_2019/html/Lin_TSM_Temporal_Shift_Module_for_Efficient_Video_Understanding_ICCV_2019_paper.html

    2019

  31. [39]

    Swin Transformer : Hierarchical Vision Transformer Using Shifted Windows

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., and Guo, B. Swin Transformer : Hierarchical Vision Transformer Using Shifted Windows . pp.\ 10012--10022, 2021. URL https://openaccess.thecvf.com/content/ICCV2021/html/Liu_Swin_Transformer_Hierarchical_Vision_Tran...

    2021

  32. [40]

    O., Mostafa, H., and Zenke, F

    Neftci, E. O., Mostafa, H., and Zenke, F. Surrogate Gradient Learning in Spiking Neural Networks : Bringing the Power of Gradient - Based Optimization to Spiking Neural Networks . IEEE Signal Processing Magazine, 36 0 (6): 0 51--63, November 2019. ISSN 1558-0792. doi:10.1109/M...

    2019

  33. [41]

    A reconfigurable on-line learning spiking neuromorphic processor comprising 256 neurons and 128K synapses

    Qiao, N., Mostafa, H., Corradi, F., Osswald, M., Stefanini, F., Sumislawska, D., and Indiveri, G. A reconfigurable on-line learning spiking neuromorphic processor comprising 256 neurons and 128K synapses. Frontiers in Neuroscience, 9, April 2015. ISSN 1662-453X. doi:10.3389/fn...

    2015

  34. [42]

    and Roy, K

    Rathi, N. and Roy, K. DIET - SNN : A Low - Latency Spiking Neural Network With Direct Input Encoding and Leakage and Threshold Optimization . IEEE Transactions on Neural Networks and Learning Systems, 34 0 (6): 0 3174--3182, June 2023. ISSN 2162-2388. doi:10.1109/TNNLS.2021.31...

    2023

  35. [43]

    P., and McGinnity, T

    Taherkhani, A., Belatreche, A., Li, Y., Cosma, G., Maguire, L. P., and McGinnity, T. M. A review of learning in biologically plausible spiking neural networks. Neural Networks, 122: 0 253--272, 2020. ISSN 0893-6080. doi:10.1016/j.neunet.2019.09.036. URL https://www.sciencedire...

    2020 doi

  36. [44]

    Turrigiano, G. G. and Nelson, S. B. Homeostatic plasticity in the developing nervous system. Nature Reviews Neuroscience, 5 0 (2): 0 97--107, February 2004. ISSN 1471-003X, 1471-0048. doi:10.1038/nrn1327. URL https://www.nature.com/articles/nrn1327

    2004 doi

  37. [45]

    A New ANN - SNN Conversion Method with High Accuracy , Low Latency and Good Robustness

    Wang, B., Cao, J., Chen, J., Feng, S., and Wang, Y. A New ANN - SNN Conversion Method with High Accuracy , Low Latency and Good Robustness . In Proceedings of IJCAI , pp.\ 3067--3075, Macau, SAR China, 2023 a . ISBN 978-1-956792-03-4. doi:10.24963/ijcai.2023/342. URL https://w...

    2023 doi

  38. [46]

    Spatial- Temporal Self - Attention for Asynchronous Spiking Neural Networks

    Wang, Y., Shi, K., Lu, C., Liu, Y., Zhang, M., and Qu, H. Spatial- Temporal Self - Attention for Asynchronous Spiking Neural Networks . In Proceedings of IJCAI , pp.\ 3085--3093, Macau, SAR China, August 2023 b . ISBN 978-1-956792-03-4. doi:10.24963/ijcai.2023/344. URL https:/...

    2023 doi

  39. [47]

    ACTION - Net : Multipath Excitation for Action Recognition

    Wang, Z., She, Q., and Smolic, A. ACTION - Net : Multipath Excitation for Action Recognition . pp.\ 13214--13223, 2021. URL https://openaccess.thecvf.com/content/CVPR2021/html/Wang_ACTION-Net_Multipath_Excitation_for_Action_Recognition_CVPR_2021_paper.html

    2021

  40. [48]

    Shift: A Zero FLOP , Zero Parameter Alternative to Spatial Convolutions

    Wu, B., Wan, A., Yue, X., Jin, P., Zhao, S., Golmant, N., Gholaminejad, A., Gonzalez, J., and Keutzer, K. Shift: A Zero FLOP , Zero Parameter Alternative to Spatial Convolutions . pp.\ 9127--9135, 2018 a . URL https://openaccess.thecvf.com/content_cvpr_2018/html/Wu_Shift_A_Zer...

    2018

  41. [49]

    Spatio- Temporal Backpropagation for Training High - Performance Spiking Neural Networks

    Wu, Y., Deng, L., Li, G., Zhu, J., and Shi, L. Spatio- Temporal Backpropagation for Training High - Performance Spiking Neural Networks . Frontiers in Neuroscience, 12: 0 323875, 2018 b . ISSN 1662-453X. URL https://www.frontiersin.org/articles/10.3389/fnins.2018.00331

    2018

  42. [50]

    Direct Training for Spiking Neural Networks : Faster , Larger , Better

    Wu, Y., Deng, L., Li, G., Zhu, J., Xie, Y., and Shi, L. Direct Training for Spiking Neural Networks : Faster , Larger , Better . Proceedings of AAAI, 33: 0 1311--1318, July 2019. doi:10.1609/aaai.v33i01.33011311. URL https://ojs.aaai.org/index.php/AAAI/article/view/3929. Number: 01

    2019 doi

  43. [51]

    K., Tang, H., and Pan, G

    Xu, Q., Li, Y., Fang, X., Shen, J., Liu, J. K., Tang, H., and Pan, G. Biologically inspired structure learning with reverse knowledge distillation for spiking neural networks, April 2023 a . URL http://arxiv.org/abs/2304.09500. arXiv:2304.09500 [cs]

    2023 arXiv

  44. [52]

    K., Tang, H., and Pan, G

    Xu, Q., Li, Y., Shen, J., Liu, J. K., Tang, H., and Pan, G. Constructing Deep Spiking Neural Networks From Artificial Neural Networks With Knowledge Distillation . In Proceedings of CVPR , pp.\ 7886--7895, June 2023 b . URL https://openaccess.thecvf.com/content/CVPR2023/html/X...

    2023

  45. [53]

    Rsnn: Recurrent spiking neural networks for dynamic spatial-temporal information processing

    Xu, Q., Fang, X., Li, Y., Shen, J., Ma, D., Xu, Y., and Pan, G. Rsnn: Recurrent spiking neural networks for dynamic spatial-temporal information processing. In Proceedings of the 32nd ACM International Conference on Multimedia, pp.\ 10602--10610, 2024

    2024

  46. [54]

    Spiking Neural Networks and Their Applications : A Review

    Yamazaki, K., Vo-Ho, V.-K., Bulsara, D., and Le, N. Spiking Neural Networks and Their Applications : A Review . Brain Sciences, 12 0 (7): 0 863, July 2022. ISSN 2076-3425. doi:10.3390/brainsci12070863. URL https://www.mdpi.com/2076-3425/12/7/863. Number: 7 Publisher: Multidisc...

    2022 doi

  47. [55]

    Attention Spiking Neural Networks

    Yao, M., Zhao, G., Zhang, H., Hu, Y., Deng, L., Tian, Y., Xu, B., and Li, G. Attention Spiking Neural Networks . IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 0 (8): 0 9393--9410, August 2023. ISSN 1939-3539. doi:10.1109/TPAMI.2023.3241201. URL https://iee...

    2023

  48. [56]

    GLIF : A Unified Gated Leaky Integrate -and- Fire Neuron for Spiking Neural Networks

    Yao, X., Li, F., Mo, Z., and Cheng, J. GLIF : A Unified Gated Leaky Integrate -and- Fire Neuron for Spiking Neural Networks . Proceedings of NeurIPS, 35: 0 32160--32171, December 2022. URL https://proceedings.neurips.cc/paper_files/paper/2022/hash/cfa8440d500a6a6867157dfd4eaff...

    2022

  49. [57]

    Temporal separation with entropy regularization for knowledge distillation in spiking neural networks

    Yu, K., Yu, C., Zhang, T., Zhao, X., Yang, S., Wang, H., Zhang, Q., and Xu, Q. Temporal separation with entropy regularization for knowledge distillation in spiking neural networks. arXiv preprint arXiv:2503.03144, 2025 a

    2025 arXiv

  50. [58]

    Fsta-snn: Frequency-based spatial-temporal attention module for spiking neural networks

    Yu, K., Zhang, T., Wang, H., and Xu, Q. Fsta-snn: Frequency-based spatial-temporal attention module for spiking neural networks. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pp.\ 22227--22235, 2025 b

    2025

  51. [59]

    S2- MLP : Spatial - Shift MLP Architecture for Vision

    Yu, T., Li, X., Cai, Y., Sun, M., and Li, P. S2- MLP : Spatial - Shift MLP Architecture for Vision . pp.\ 297--306, 2022. URL https://openaccess.thecvf.com/content/WACV2022/html/Yu_S2-MLP_Spatial-Shift_MLP_Architecture_for_Vision_WACV_2022_paper.html

    2022

  52. [60]

    Da-lif: Dual adaptive leaky integrate-and-fire model for deep spiking neural networks

    Zhang, T., Yu, K., Zhang, J., and Wang, H. Da-lif: Dual adaptive leaky integrate-and-fire model for deep spiking neural networks. In ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp.\ 1--5. IEEE, 2025 a

    2025

  53. [61]

    Staa-snn: Spatial-temporal attention aggregator for spiking neural networks

    Zhang, T., Yu, K., Zhong, X., Wang, H., Xu, Q., and Zhang, Q. Staa-snn: Spatial-temporal attention aggregator for spiking neural networks. arXiv preprint arXiv:2503.02689, 2025 b

    2025 arXiv

  54. [62]

    Going Deeper With Directly - Trained Larger Spiking Neural Networks

    Zheng, H., Wu, Y., Deng, L., Hu, Y., and Li, G. Going Deeper With Directly - Trained Larger Spiking Neural Networks . Proceedings of AAAI, 35: 0 11062--11070, May 2021. doi:10.1609/aaai.v35i12.17320. URL https://ojs.aaai.org/index.php/AAAI/article/view/17320. Number: 12

    2021 doi

  55. [63]

    Spike- Based Motion Estimation for Object Tracking Through Bio - Inspired Unsupervised Learning

    Zheng, Y., Yu, Z., Wang, S., and Huang, T. Spike- Based Motion Estimation for Object Tracking Through Bio - Inspired Unsupervised Learning . IEEE Transactions on Image Processing, 32: 0 335--349, 2023. ISSN 1941-0042. doi:10.1109/TIP.2022.3228168. URL https://ieeexplore.ieee.o...

    2023

  56. [64]

    Spikformer: When Spiking Neural Network Meets Transformer

    Zhou, Z., Zhu, Y., He, C., Wang, Y., Yan, S., Tian, Y., and Yuan, L. Spikformer: When Spiking Neural Network Meets Transformer . In ICLR , September 2022. URL https://openreview.net/forum?id=frE4fUwz_h

    2022

Pith tools

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