Pith. sign in

REVIEW 3 major objections 5 minor 44 references

TDFormer: A Top-Down Attention-Controlled Spiking Transformer

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 A top-down feedback pathway lifts spiking transformer accuracy to 86.83% on ImageNet.

desk verdict A clean top-down feedback module for spiking transformers with consistent small gains, but the ImageNet SOTA number is built on an unmatched fine-tuning baseline and the vanishing-gradient theorem is asserted rather than proved. read the letter →

arxiv 2505.15840 v2 pith:V53SPRNP submitted 2025-05-17 cs.NE cs.AIcs.CV

Zizheng Zhu , Yingchao Yu , Zeqi Zheng , Zhaofei Yu , Yaochu Jin This is my paper · ORCID
classification cs.NEcs.AIcs.CV MSC 68T07
keywords spikingneuralnetworkstransformertop-downfeedbacktemporalgradientvanishingspike-basedself-attentionImageNetclassificationLIFneuronmutualinformationacrosstimesteps
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

TDFormer argues that existing transformer-based spiking neural networks are temporally impoverished: each time step runs like a separate subnetwork whose only link to the past is the membrane potential, so features across time carry little mutual information and temporal gradients fade. The paper's proposed fix is a top-down attention-controlled feedback pathway that carries high-order representations from a later time step back to modulate an earlier one through a learned differentiable function $\varphi_\theta(S_l(t))$. On the forward pass this raises mutual information across time steps; on the backward pass the paper proves that the same feedback term makes the per-step temporal sensitivity $\epsilon^l(t)_{jj}$ nonzero inside the firing-threshold interval, where a conventional LIF neuron has sensitivity zero. Empirically, the model reaches 86.83% top-1 accuracy on ImageNet at four time steps, with minimal added parameters and power over its QKFormer baseline, which the paper reports as the best SNN number on that benchmark. The claim matters because it points to a structural, biologically inspired fix for temporal representation in SNNs rather than a training-only workaround.

What carries the argument

The load-bearing machinery is the TDAC top-down pathway, made of a control module (CM) and a processing module (PM), wrapped around a transformer backbone built on leaky integrate-and-fire (LIF) neurons. CM concatenates the bottom-up spikes of the current stage with the top-down spikes of the previous stage to form the key, while queries and values come from bottom-up features, so attention is controlled by higher-order temporal context. PM computes a spatial attention map $M_{\rm spatial}=\operatorname{clamp}(\sum_c W_c X_{t,n,c}, b, a)$, multiplies it into the spiking feature map, and re-spikes, producing the feedback signal $S^{(n)}_{td}$ and the additive differentiable term $\varphi_\theta(S_l(t))$ in the membrane dynamics. Proposition 4.1 bounds the variance of the PM output $X \odot M_{\rm spatial}$ under a Bernoulli firing-rate model, and Proposition B.4 gives mean and variance approximations that justify the recommended setting $b=0$, $a\in[1,2]$. The whole design is expressed as a sequence of $N$ sub-networks, each with its own classification loss, so the feedback path doubles as a temporal hierarchy.

What would settle it

Train TDFormer and its baseline to convergence, then freeze the model and evaluate $\partial\varphi_\theta(S_l(t))/\partial S_l(t)$ for all membrane potentials in the band $\vartheta/2 < H^l_j(t) < 3\vartheta/2$ across layers and time steps. If the mean absolute value of that derivative is close to zero, or below $1-1/\tau$, in the trained network, then the claimed vanishing-gradient relief does not hold for the learned feedback path. A complementary check is to compare the temporal-gradient norm through BPTT with the feedback term included versus ablated, at matched accuracy.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that adding a top-down feedback term to the membrane potential update changes both what an SNN represents over time and how it learns over time. In the forward direction, the control module CM folds the previous stage's top-down spikes into the key computation of spike-based self-attention, and the processing module PM produces a spatially gated, clamped spiking feature map, so later high-order representations steer earlier low-order processing. In the backward direction, the paper proves (Theorem 4.3) that with a rectangular surrogate gradient the per-step sensitivity of a conventional leaky integrate-and-fire (LIF) neuron is $\epsilon^l(t)_{jj}=0$ when the membrane potential lies in the interval $(\vartheta/2, 3\vartheta/2)$ and $1-1/\tau$ otherwise, while with the feedback path it is $\partial\varphi_\theta(S_l(t))/\partial S_l(t)$ in that interval, which the paper argues is nonzero and can exceed $1-1/\tau$. That is the mechanism claimed to alleviate temporal vanishing gradients. The measured consequence is consistent gains over three spiking transformer baselines on CIFAR-10, CIFAR-100, CIFAR10-DVS, DVS128 Gesture, and ImageNet, with the ImageNet result at 86.83%.

Load-bearing premise

The proof that feedback removes temporal vanishing gradients assumes the learned feedback function $\varphi_\theta$ has a derivative with respect to the spike that is never zero inside the firing-threshold band, yet the paper never specifies $\varphi_\theta$ or proves that nonzero property.

Editorial extensions

If this is right

  • If the central claim is right, transformer-based SNNs can reach ANN-level accuracy on ImageNet (86.83% at four time steps), narrowing the gap to non-spiking vision transformers on the benchmark.
  • The temporal-gradient result implies that training deeper or longer-time-step SNNs should suffer less from gradient decay along time, making the time dimension a more usable resource for network design.
  • The feedback path's low overhead means the accuracy gain does not come at the cost of the energy advantage that motivates SNNs.
  • Raising mutual information between time steps suggests the model integrates temporal context rather than just averaging spike rates, which should improve tasks where ordering or accumulation matters.
  • Weighted supervision of every sub-network output gives earlier stages their own learning signal, consistent with the claim that temporal receptive field grows along the feedback hierarchy.

Reading between the lines

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

  • Editorial: the same control-module pattern could be tested on CNN-based SNNs by replacing the attention-based key fusion with a convolutional feedback mixer; the paper lists transformer-only generality as a limitation, so this is a natural next experiment.
  • Editorial: if the gradient mechanism is real, it should appear as a larger temporal gradient norm early in training for TDFormer than for its baseline; measuring that norm at matched loss would separate the gradient benefit from the mutual-information benefit.
  • Editorial: the variance bound assumes independent, low-rate Bernoulli spikes, so under correlated or high-rate firing the recommended clamp setting $b=0$, $a\in[1,2]$ may need re-tuning; the paper does not report that stress test.
  • Editorial: because the feedback term is additive in the membrane update and differentiable, the same construction could be dropped into other time-rolled architectures, such as recurrent or equilibrium models, as a generic temporal-context gate.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes TDFormer, a modification of three transformer-based spiking neural network backbones (SpikformerV1, Spike-driven TransformerV1, and QKFormer) that adds a top-down feedback pathway called TDAC. The pathway consists of a control module (CM), which mixes bottom-up and top-down signals to modulate attention, and a processing module (PM), which generates the feedback signal from higher-level representations. The authors report consistent accuracy improvements on CIFAR-10, CIFAR-100, CIFAR10-DVS, DVS128 Gesture, and ImageNet, where they claim a new state of the art of 86.83% top-1 accuracy with minimal parameter and energy overhead. The paper also presents two theoretical contributions: a variance bound for the PM attention output that motivates a clamp operation, and a theorem claiming that the feedback term alleviates temporal vanishing gradients in SNNs.

Significance. If the reported ImageNet result is trustworthy, the paper makes a significant empirical contribution: it introduces a structurally new mechanism for temporal information flow in transformer-based SNNs, demonstrates accuracy close to ANN-level on ImageNet, and supports the main claims with ablations across CM/PM variants, multiple backbones, multiple seeds, corruption robustness experiments, and energy-consumption estimates. The theoretical analysis is less central but could add value if made rigorous. However, the headline ImageNet claim currently rests on an unmatched fine-tuning comparison, and the vanishing-gradient theorem is asserted rather than proved for the actual feedback modules. These issues are load-bearing for the paper's main claims.

major comments (3)
  1. [Appendix A.1, Table 1] The ImageNet comparison is not matched. Appendix A.1 states that for QKFormer the authors 'fine-tuned the pretrained network with a base learning rate of 2e-5 for 15 epochs' to obtain the TDFormer results, yet Table 1 reports QKFormer baseline accuracies (84.22, 85.20, 85.65) that appear to be the published numbers, with no 'QKFormer(ours)' row re-run under the same 15-epoch fine-tuning recipe. The reported +1.15 to +1.18 gains therefore conflate the effect of the TDAC module with the effect of continued training of a pretrained checkpoint at a low learning rate. The central SOTA claim of 86.83% is not established unless the QKFormer baseline is re-run under the identical fine-tuning protocol, or TDFormer is compared against QKFormer trained from scratch under the original protocol.
  2. [Theorem 4.3, Appendix B.3, Eq. (108)] The vanishing-gradient theorem is not supported as stated. The proof assumes a differentiable feedback function φ_θ, but the manuscript never specifies φ_θ analytically or proves that ∂φ_θ(S_l(t))/∂S_l(t) is nonzero inside the firing-threshold interval; the sentence after Eq. (108) merely asserts this. In addition, there is a factor inconsistency: inside the interval 1/2ϑ < H < 3/2ϑ the rectangular surrogate gives ∂S/∂H = 1/ϑ, so combining Eqs. (106) and (107) yields ϵ_l(t)_jj = (1/ϑ) · ∂φ_θ/∂S_l(t), not ∂φ_θ/∂S_l(t) as written in Eq. (108). The theorem and its proof need to be corrected, and the nonzero-derivative condition must either be proved for the actual PM/CM modules or explicitly stated as an assumption with empirical verification.
  3. [Proposition B.4, Appendix B.1] The variance analysis that motivates the clamp operation and the hyperparameter recommendation b=0, a∈[1,2] relies on strong, partly unstated assumptions: asymptotic independence of X and M (Lemma B.3), b≈0, a≥1, small firing rate f, and the approximation μ≈0 with exponential tail terms dropped. The errors in these approximations are not controlled, so the recommendation is not rigorously grounded. Please state these assumptions in the main text and provide empirical validation, for example by reporting measured firing rates and the variance of M against the predicted bounds.
minor comments (5)
  1. [Section 5.1, Table 1] The text says 'increasing the model size by merely 0.02 million parameters and 0.59 millijoules of power consumption,' but the first TDFormer row versus QKFormer shows +0.59 million parameters and +0.02 mJ; the two quantities are swapped.
  2. [Section 5.2, Table 2] The text refers to 'the exception of Spikformer-4-384 (V1) at time step 6,' but Table 2 only shows time steps 2 and 4; the exception appears in Table 5 for CIFAR-100/T=6, so the cross-reference should be corrected.
  3. [Section 5.3, Appendix A.2] Section 5.3 says the robustness evaluation uses CIFAR-10C with 15 corruption types, while Appendix A.2 and the caption of Table 7 state 19 corruption types; the number should be consistent.
  4. [Equations (10)-(14)] The notation for shapes is confusing: S_bu is given shape T×H×W×C while H_n and S_td have shape T×N×C, even though the control module concatenates these along the channel dimension; the manuscript should clarify that N = H·W (or otherwise define the reshaping explicitly).
  5. [Appendix B.3, Eq. (107)] The derivation of Eq. (107) is missing the factor 1/ϑ from ∂S/∂H when simplifying the feedback term; this is related to the major comment on Eq. (108) and should be fixed consistently.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical claims are tested against external baselines, and the one unsupported theoretical assertion is a missing proof, not a self-referential derivation.

full rationale

The central empirical claims are measured accuracies on standard benchmarks (ImageNet, CIFAR, DVS) compared with external prior models, not quantities derived from fitted parameters, so there is no construction-level circularity in the headline results. The theoretical derivations (Proposition 4.1, Proposition B.4, and the gradient formulas in Equations 103-108) proceed from explicitly stated assumptions and do not reduce to their own inputs. The only self-citation, reference [21] in related work, is background context and is not load-bearing. Two non-circular weaknesses should be noted. First, Appendix A.1 states that ImageNet TDFormer models were fine-tuned from pretrained QKFormer checkpoints for 15 epochs with a base learning rate of 2e-5, while Table 1 lists QKFormer baseline rows that appear to be the published numbers rather than QKFormer re-run under the same fine-tuning protocol; this is an experimental confound that may inflate the reported gains, but it is not a case of a prediction being equivalent to its inputs. Second, Appendix B.3 asserts after Equation (108) that 'in training, ∂φθ(Sl(t))/∂Sl(t) is not possible to be zero,' and the vanishing-gradient claim of Theorem 4.3 rests on this assertion, yet φθ is never specified analytically and no proof of nonzero derivative is given. This is an omitted proof, a correctness risk rather than circularity. Under the stated criteria, a non-finding is appropriate: the derivation chain is not circular.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The ledger shows a moderate axiomatic load. The architecture introduces no new physical entities, but the theoretical contribution rests on several untested stochastic independence assumptions and on an unproven property of the feedback function. The most consequential unspecified parameters are the loss weights alpha_n and the actual firing threshold/tau settings, which are needed to reproduce the training.

free parameters (4)
  • clamp bounds a and b = b = 0, a in [1, 2] (recommended range)
    The clamp operation in the spatial processing module uses bounds a and b; the paper recommends b=0 and a in [1,2] based on the variance analysis, but does not report a systematic sweep across datasets or show sensitivity to these choices.
  • loss weights alpha_n = not specified numerically
    The loss is a weighted sum over N sub-network outputs with alpha_n summing to 1, but the actual alpha values used are not reported, making the training procedure underspecified.
  • spike threshold and time constant tau = not reported
    The analysis assumes tau is close to 1 and that firing rates are low, but the actual tau and threshold values used in experiments are not listed. The variance bound and gradient conclusions depend on these.
  • PM variant choice = v1 as default, with v2/v3/v4 explored
    The four PM variants are architectural choices selected by validation accuracy; the default variant is chosen after comparing variants, which is a model selection step that contributes to the headline results.
assumptions (5)
  • domain assumption Spike activations can be treated as independent Bernoulli random variables with a common firing rate f.
    Used in Lemmas B.2 and B.3 and Proposition 4.1 to derive the variance bound and the asymptotic independence of X and M. Real spiking activations are correlated across channels and time, so the bound is an approximation.
  • ad hoc to paper The feedback function phi_theta(S_l(t)) is differentiable and has nonzero derivative inside the firing-threshold interval.
    Theorem 4.3's conclusion that the feedback gradient replaces zero with a nonzero value depends on this property, which is asserted without proof or an explicit form for phi.
  • domain assumption The rectangular surrogate gradient is the correct model for backpropagation through spikes.
    The gradient analysis in Section B.3 and Theorem 4.3 uses the rectangular surrogate, following earlier work. Different surrogate functions would change the numerical conclusions.
  • standard math Standard results about characteristic functions and the Central Limit Theorem apply to sums of independent Bernoulli variables.
    Used implicitly in Lemmas B.2 and B.3 to show that the token mixing sum becomes Gaussian as C grows.
  • domain assumption Accuracy on ImageNet after fine-tuning a pretrained backbone for 15 epochs reflects the quality of the proposed module.
    The ImageNet protocol is a short fine-tuning run of a pretrained QKFormer. This assumes the comparison is fair relative to baselines that used full training schedules, and that the fine-tuning does not heavily depend on the pretrained checkpoint.
invented entities (1)
  • TDAC top-down pathway with Control Module (CM) and Processing Module (PM)
    purpose: Adds explicit cross-time-step feedback of high-level features to modulate attention in later processing stages.
    This is a new architectural construct. Its only evidence base is the experiments in this paper; there is no external falsifiable prediction outside the reported accuracy numbers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TDFormer: A Top-Down Attention-Controlled Spiking Transformer." pith.science (2026). https://pith.science/paper/V53SPRNP

@misc{pith2026250515840,
  author       = {Pith},
  title        = {Pith review of: TDFormer: A Top-Down Attention-Controlled Spiking Transformer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V53SPRNP}},
  note         = {Machine review of arXiv:2505.15840}
}
read the original abstract

Traditional spiking neural networks (SNNs) can be viewed as a combination of multiple subnetworks with each running for one time step, where the parameters are shared, and the membrane potential serves as the only information link between them. However, the implicit nature of the membrane potential limits its ability to effectively represent temporal information. As a result, each time step cannot fully leverage information from previous time steps, seriously limiting the model's performance. Inspired by the top-down mechanism in the brain, we introduce TDFormer, a novel model with a top-down feedback structure that functions hierarchically and leverages high-order representations from earlier time steps to modulate the processing of low-order information at later stages. The feedback structure plays a role from two perspectives: 1) During forward propagation, our model increases the mutual information across time steps, indicating that richer temporal information is being transmitted and integrated in different time steps. 2) During backward propagation, we theoretically prove that the feedback structure alleviates the problem of vanishing gradients along the time dimension. We find that these mechanisms together significantly and consistently improve the model performance on multiple datasets. In particular, our model achieves state-of-the-art performance on ImageNet with an accuracy of 86.83%.

Figures

Figures reproduced from arXiv: 2505.15840 by the authors.

Figure 1
Figure 1. Visualization of mutual information matrices of features across time steps on ImageNet. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the TDFormer architecture. (a) Overall design inspired by top-down pathways [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. This is the histogram of the gradient of the surrogate function for LIF neurons in the [PITH_FULL_IMAGE:figures/full_fig_p024_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visualization of CIFAR-10C. This figure showcases 19 columns corresponding to 19 [PITH_FULL_IMAGE:figures/full_fig_p025_4.png]
Figure 5
Figure 5. Figure 5: Visualization of ImageNet-C. This figure showcases 19 columns corresponding to 19 [PITH_FULL_IMAGE:figures/full_fig_p026_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 19 canonical work pages

  1. [1]

    A comprehensive review of spiking neural networks: Interpretation, optimization, efficiency, and best practices

    Kai Malcolm and Josue Casco-Rodriguez. A comprehensive review of spiking neural networks: Interpretation, optimization, efficiency, and best practices. arXiv preprint arXiv:2303.10780, 2023

    arXiv 2023

  2. [2]

    Spikformer: When spiking neural network meets transformer

    Zhaokun Zhou, Yuesheng Zhu, Chao He, Yaowei Wang, Shuicheng YAN, Yonghong Tian, and Li Yuan. Spikformer: When spiking neural network meets transformer. In The Eleventh International Conference on Learning Representations, 2023

    2023

  3. [3]

    Qkformer: Hierarchical spiking transformer using qk attention

    Chenlin Zhou, Han Zhang, Zhaokun Zhou, Liutao Yu, Liwei Huang, Xiaopeng Fan, Li Yuan, Zhengyu Ma, Huihui Zhou, and Yonghong Tian. Qkformer: Hierarchical spiking transformer using qk attention. arXiv preprint arXiv:2403.16552, 2024

    arXiv 2024

  4. [4]

    Spikingformer: Spike-driven residual learning for transformer-based spiking neural network

    Chenlin Zhou, Liutao Yu, Zhaokun Zhou, Zhengyu Ma, Han Zhang, Huihui Zhou, and Yonghong Tian. Spikingformer: Spike-driven residual learning for transformer-based spiking neural network. arXiv preprint arXiv:2304.11954, 2023

    arXiv 2023

  5. [5]

    Spike- driven transformer

    Man Yao, JiaKui Hu, Zhaokun Zhou, Li Yuan, Yonghong Tian, Bo XU, and Guoqi Li. Spike- driven transformer. In Thirty-seventh Conference on Neural Information Processing Systems, 2023

    work page 2023

  6. [6]

    Spike-driven transformer v2: Meta spiking neural network architecture inspiring the design of next-generation neuromorphic chips

    Man Yao, JiaKui Hu, Tianxiang Hu, Yifan Xu, Zhaokun Zhou, Yonghong Tian, Bo XU, and Guoqi Li. Spike-driven transformer v2: Meta spiking neural network architecture inspiring the design of next-generation neuromorphic chips. In The Twelfth International Conference on Learning Representations, 2024

    work page 2024

  7. [7]

    Direct training for spiking neural networks: Faster, larger, better

    Yujie Wu, Lei Deng, Guoqi Li, Jun Zhu, Yuan Xie, and Luping Shi. Direct training for spiking neural networks: Faster, larger, better. In Proceedings of the AAAI conference on artificial intelligence, volume 33, pages 1311–1318, 2019

    work page 2019

  8. [8]

    Rethinking spiking neural networks from an ensemble learning perspective

    Yongqi Ding, Lin Zuo, Mengmeng Jing, Pei He, and Hanpu Deng. Rethinking spiking neural networks from an ensemble learning perspective. arXiv preprint arXiv:2502.14218, 2025

    arXiv 2025

Show all 44 references
  1. [9]

    Towards memory-and time-efficient backpropagation for training spiking neural networks

    Qingyan Meng, Mingqing Xiao, Shen Yan, Yisen Wang, Zhouchen Lin, and Zhi-Quan Luo. Towards memory-and time-efficient backpropagation for training spiking neural networks. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 6166–6176, 2023

    2023

  2. [10]

    Rethinking the membrane dynamics and optimization objectives of spiking neural networks

    Hangchi Shen, Qian Zheng, Huamin Wang, and Gang Pan. Rethinking the membrane dynamics and optimization objectives of spiking neural networks. Advances in Neural Information Processing Systems, 37:92697–92720, 2024

    2024

  3. [11]

    Deeptage: Deep temporal-aligned gradient enhancement for optimizing spiking neural networks

    Wei Liu, Li Yang, Mingxuan Zhao, Shuxun Wang, Jin Gao, Wenjuan Li, Bing Li, and Weiming Hu. Deeptage: Deep temporal-aligned gradient enhancement for optimizing spiking neural networks. In The Thirteenth International Conference on Learning Representations, 2025

    2025

  4. [12]

    Clif: Complementary leaky integrate-and-fire neuron for spiking neural networks

    Yulong Huang, Xiaopeng Lin, Hongwei Ren, Haotian Fu, Yue Zhou, Zunchang Liu, Biao Pan, and Bojun Cheng. Clif: Complementary leaky integrate-and-fire neuron for spiking neural networks. arXiv preprint arXiv:2402.04663, 2024

    2024 arXiv

  5. [13]

    Spiking transformer with spatial-temporal attention

    Donghyun Lee, Yuhang Li, Youngeun Kim, Shiting Xiao, and Priyadarshini Panda. Spiking transformer with spatial-temporal attention. arXiv preprint arXiv:2409.19764, 2024. 10

    2024 arXiv

  6. [14]

    Top-down influences on visual processing

    Charles D Gilbert and Wu Li. Top-down influences on visual processing. Nature reviews neuroscience, 14(5):350–363, 2013

    2013

  7. [15]

    Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices

    Timothy J Buschman and Earl K Miller. Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. science, 315(5820):1860–1862, 2007

    2007

  8. [16]

    The normalization model of attention.Neuron, 61(2):168– 185, 2009

    John H Reynolds and David J Heeger. The normalization model of attention.Neuron, 61(2):168– 185, 2009

    2009

  9. [17]

    A common network of functional areas for attention and eye movements

    Maurizio Corbetta, Erbil Akbudak, Thomas E Conturo, Abraham Z Snyder, John M Ollinger, Heather A Drury, Martin R Linenweber, Steven E Petersen, Marcus E Raichle, David C Van Essen, et al. A common network of functional areas for attention and eye movements. Neuron, 21(4):761–773, 1998

    1998

  10. [18]

    Spikformer v2: Join the high accuracy club on imagenet with an snn ticket

    Zhaokun Zhou, Kaiwei Che, Wei Fang, Keyu Tian, Yuesheng Zhu, Shuicheng Yan, Yonghong Tian, and Li Yuan. Spikformer v2: Join the high accuracy club on imagenet with an snn ticket. arXiv preprint arXiv:2401.02020, 2024

    2024 arXiv

  11. [19]

    Spikegpt: Generative pre-trained language model with spiking neural networks

    Rui-Jie Zhu, Qihang Zhao, Guoqi Li, and Jason K Eshraghian. Spikegpt: Generative pre-trained language model with spiking neural networks. arXiv preprint arXiv:2302.13939, 2023

    2023 arXiv

  12. [20]

    Sglformer: Spiking global-local-fusion transformer with high performance

    Han Zhang, Chenlin Zhou, Liutao Yu, Liwei Huang, Zhengyu Ma, Xiaopeng Fan, Huihui Zhou, and Yonghong Tian. Sglformer: Spiking global-local-fusion transformer with high performance. Frontiers in Neuroscience, 18:1371290, 2024

    2024

  13. [21]

    Object recognition using a bio-inspired neuron model with bottom-up and top-down pathways

    Yuhua Zheng, Yan Meng, and Yaochu Jin. Object recognition using a bio-inspired neuron model with bottom-up and top-down pathways. Neurocomputing, 74(17):3158–3169, 2011

    2011

  14. [22]

    Bottom-up and top-down attention for image captioning and visual question answering

    Peter Anderson, Xiaodong He, Chris Buehler, Damien Teney, Mark Johnson, Stephen Gould, and Lei Zhang. Bottom-up and top-down attention for image captioning and visual question answering. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 60...

    2018

  15. [23]

    Top-down visual attention from analysis by synthesis

    Baifeng Shi, Trevor Darrell, and Xin Wang. Top-down visual attention from analysis by synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2102–2112, 2023

    2023

  16. [24]

    Top-down network combines back-propagation with attention

    Roy Abel and Shimon Ullman. Top-down network combines back-propagation with attention. arXiv preprint arXiv:2306.02415, 2023

    2023 arXiv

  17. [25]

    Cambridge University Press, 2014

    Wulfram Gerstner, Werner M Kistler, Richard Naud, and Liam Paninski.Neuronal dynamics: From single neurons to networks and models of cognition. Cambridge University Press, 2014

    2014

  18. [26]

    Attention is all you need

    A Vaswani. Attention is all you need. Advances in Neural Information Processing Systems, 2017

    2017

  19. [27]

    Metaformer is actually what you need for vision

    Weihao Yu, Mi Luo, Pan Zhou, Chenyang Si, Yichen Zhou, Xinchao Wang, Jiashi Feng, and Shuicheng Yan. Metaformer is actually what you need for vision. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10819–10829, 2022

    2022

  20. [28]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv...

    2010 arXiv

  21. [29]

    Training data-efficient image transformers & distillation through attention

    Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Hervé Jégou. Training data-efficient image transformers & distillation through attention. In International conference on machine learning, pages 10347–10357. PMLR, 2021

    2021

  22. [30]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 11

    2021

  23. [31]

    Scaling spike-driven transformer with efficient spike firing approximation training

    Man Yao, Xuerui Qiu, Tianxiang Hu, Jiakui Hu, Yuhong Chou, Keyu Tian, Jianxing Liao, Luziwei Leng, Bo Xu, and Guoqi Li. Scaling spike-driven transformer with efficient spike firing approximation training. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    2025

  24. [32]

    Temporal efficient training of spiking neural network via gradient re-weighting

    Shikuang Deng, Yuhang Li, Shanghang Zhang, and Shi Gu. Temporal efficient training of spiking neural network via gradient re-weighting. arXiv preprint arXiv:2202.11946, 2022

    2022 arXiv

  25. [33]

    Going deeper with directly-trained larger spiking neural networks

    Hanle Zheng, Yujie Wu, Lei Deng, Yifan Hu, and Guoqi Li. Going deeper with directly-trained larger spiking neural networks. In Proceedings of the AAAI conference on artificial intelligence, volume 35, pages 11062–11070, 2021

    2021

  26. [34]

    Recdis-snn: Rectifying membrane potential distribution for directly training spiking neural net- works

    Yufei Guo, Xinyi Tong, Yuanpei Chen, Liwen Zhang, Xiaode Liu, Zhe Ma, and Xuhui Huang. Recdis-snn: Rectifying membrane potential distribution for directly training spiking neural net- works. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, ...

    2022

  27. [35]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky. Learning multiple layers of features from tiny images. Technical report, University of Toronto, 2009

    2009

  28. [36]

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

    Hongmin Li, Hanchao Liu, Xiangyang Ji, Guoqi Li, and Luping Shi. Cifar10-dvs: an event- stream dataset for object classification. Frontiers in neuroscience, 11:309, 2017

    2017

  29. [37]

    A low power, fully event-based gesture recognition system

    Arnon Amir, Brian Taba, David Berg, Timothy Melano, Jeffrey McKinstry, Carmelo Di Nolfo, Tapan Nayak, Alexander Andreopoulos, Guillaume Garreau, Marcela Mendoza, et al. A low power, fully event-based gesture recognition system. In Proceedings of the IEEE conference on computer...

    2017

  30. [38]

    Imagenet: A large- scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large- scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    2009

  31. [39]

    Benchmarking neural network robustness to common corruptions and perturbations

    Dan Hendrycks and Thomas Dietterich. Benchmarking neural network robustness to common corruptions and perturbations. arXiv preprint arXiv:1903.12261, 2019

    1903 arXiv

  32. [40]

    Training high-performance low-latency spiking neural networks by differentiation on spike representation

    Qingyan Meng, Mingqing Xiao, Shen Yan, Yisen Wang, Zhouchen Lin, and Zhi-Quan Luo. Training high-performance low-latency spiking neural networks by differentiation on spike representation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pa...

    2022

  33. [41]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017

    2017 arXiv

  34. [42]

    Integer-valued training and spike-driven inference spiking neural network for high-performance and energy-efficient object detection

    Xinhao Luo, Man Yao, Yuhong Chou, Bo Xu, and Guoqi Li. Integer-valued training and spike-driven inference spiking neural network for high-performance and energy-efficient object detection. In European Conference on Computer Vision, pages 253–272. Springer, 2024

    2024

  35. [43]

    Beyond classification: Directly training spiking neural networks for semantic segmentation

    Youngeun Kim, Joshua Chough, and Priyadarshini Panda. Beyond classification: Directly training spiking neural networks for semantic segmentation. Neuromorphic Computing and Engineering, 2(4):044015, 2022

    2022

  36. [44]

    Spiking convolutional neural networks for text classification

    Changze Lv, Jianhan Xu, and Xiaoqing Zheng. Spiking convolutional neural networks for text classification. arXiv preprint arXiv:2406.19230, 2024. 12 A Implementation Details A.1 Training Protocols We adopted the following training protocols: • Spike Generation: We used a rate-...

    2024 arXiv

Pith tools

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