Pith. sign in

REVIEW 4 major objections 7 minor 37 references

Self-Attentive Spatio-Temporal Calibration for Precise Intermediate Layer Matching in ANN-to-SNN Distillation

T0 review · 4 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Learned layer matching lets distilled spiking networks beat their ANN teachers on CIFAR-10 and CIFAR-100.

desk verdict Strong empirical results for attention-based layer matching in ANN-to-SNN distillation, but the 'first time' claim is contradicted by the paper's own table and the semantic-alignment story is not independently validated. read the letter →

arxiv 2501.08049 v1 pith:MHP275ZJ submitted 2025-01-14 cs.AI cs.CVcs.LG

Di Hong , Yueming Wang This is my paper · ORCID
classification cs.AIcs.CVcs.LG
keywords spikingneuralnetworksknowledgedistillationself-attentionANN-to-SNNconversionintermediatelayermatchingspatio-temporalcalibrationneuromorphicdatasetslow-latencyinference
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

ANN-to-SNN knowledge distillation usually either distills only the teacher's class labels or forces each student layer to imitate a fixed teacher layer, which the authors argue mismatches semantic content across architectures and across spike time steps. This paper proposes self-attentive spatio-temporal calibration (SASTC), which lets each SNN layer at each time step attend to all candidate ANN layers with learned weights, so the student receives a soft, semantically weighted combination of teacher features. On this basis the paper reports SNN accuracies of 95.12% on CIFAR-10 and 79.40% on CIFAR-100 at two time steps, 68.69% on ImageNet at four time steps, and state-of-the-art results on DVS-Gesture and DVS-CIFAR10. The headline claim is that this is the first time SNNs outperform their ANN teachers on both CIFAR-10 and CIFAR-100.

What carries the argument

The machinery is a batch-level pairwise similarity matrix: $A_{s_l}^t=R(f_{s_l}^t)R(f_{s_l}^t)^T$ for each SNN layer at time $t$ and $A_{a_l}=R(f_{a_l})R(f_{a_l})^T$ for each ANN layer, where $R$ flattens feature maps and $A$ is $b\times b$. Each matrix is projected by a shared MLP into a query or key vector, and softmax attention over teacher layers yields weights $\eta_{t,i}$, which weight the MSE between each projected student feature map and each teacher feature map in the total loss $\mathcal{L}_{\text{total}}=\mathcal{L}_{KD}+\beta\mathcal{L}_{SASTC}$. The similarity matrices make semantic proximity computable at manageable memory cost and give the attention mechanism an instance-level signal that changes with both layer depth and time step; the projected query/key form is what converts that signal into a differentiable layer-matching policy.

What would settle it

Train SASTC on CIFAR-100 with the attention weights frozen to uniform (or random) values while keeping everything else identical; if accuracy stays close to the reported 79.40%, the learned alignment is not what drives the improvement. Alternatively, compare the learned teacher-layer choices against an independent semantic measure such as layer-wise linear-probe accuracy on held-out classes; attention weights that systematically select semantically unrelated layers would falsify the claim that the calibration is semantic.

Watch

Extended reading notes

Core claim

The central claim is that the performance gap in ANN-to-SNN distillation is not mainly a capacity problem but an alignment problem: a given SNN layer at a given time step does not necessarily carry the same semantic content as the ANN layer at the same depth, so fixed layer-wise imitation can actively hurt. SASTC replaces fixed associations with learned attention over the full set of teacher layers for every student layer and every time step. The authors introduce the Spatio-Temporal Mismatch (STM) score, the average Euclidean distance between the pairwise similarity matrices of associated ANN/SNN feature maps, and report that baseline, KD, and feature-distillation methods barely lower it or raise it, while SASTC lowers it and produces consistent accuracy gains. They take these results to show that autonomous spatio-temporal calibration transfers relevant semantic information and avoids the negative regularization seen with hand-specified layer pairs.

Load-bearing premise

The load-bearing premise is that two layers are semantically alike exactly when their per-instance pairwise similarity matrices are close, so the attention weights inherit whatever noise or task-dependence lives in that proxy.

Editorial extensions

If this is right

  • At two time steps the distilled SNN can already beat its ANN teacher on CIFAR-10 and CIFAR-100, so the usual latency/accuracy trade-off of spiking networks is shifted.
  • Heterogeneous teacher–student pairs (for example VGG teacher with ResNet student, or wider vs narrower models) can be distilled without hand-designing layer correspondences.
  • The reported gains on DVS-Gesture and DVS-CIFAR10 indicate the temporal alignment component transfers to neuromorphic, event-stream data.
  • The same SASTC-trained student retains more accuracy than teacher or baseline under 10–50% label noise and with 25–75% of training data, suggesting the soft alignment acts as a regularizer.

Reading between the lines

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

  • One testable extension would be to replace the similarity-matrix proxy with a task-independent semantic measure (for example, layer-wise probing accuracy or activation-pattern overlap) and check whether the learned attention weights agree; this would separate semantic alignment from optimization convenience.
  • Because the mechanism only needs pairwise similarity matrices, the same calibration could be applied to other teacher–student transfers where architectures differ in depth, such as distilling a large Transformer into a small one, with time steps replaced by layer groups.
  • A natural ablation the paper does not report is learning one fixed soft association (shared across time steps) rather than per-time-step weights; that comparison would isolate the temporal component of the claimed spatio-temporal calibration.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes SASTC (Self-Attentive Spatio-Temporal Calibration), a method for ANN-to-SNN knowledge distillation that uses self-attention to select teacher-student layer pairs at each time step. The selection is based on the proximity of Gram matrices of feature maps, and the resulting weighted MSE regularizer is combined with standard KD and CE losses. The authors report strong accuracy results on CIFAR-10 (95.12%), CIFAR-100 (79.40%), ImageNet (68.69%), DVS-Gesture (97.92%), and DVS-CIFAR10 (83.60%), and claim that this is the first time SNNs outperform ANNs on both CIFAR-10 and CIFAR-100. The paper also provides negative-regularization experiments, an STM score analysis, noisy-label and few-shot results, and efficiency comparisons.

Significance. If the reported accuracies are reproducible, SASTC is a practically strong contribution to low-latency SNN training, achieving state-of-the-art results with only 2-4 time steps on static datasets and competitive results on neuromorphic datasets. The proposed idea of soft, attention-based layer matching is a reasonable extension of existing layer-wise distillation and could be useful beyond ANN-to-SNN transfer. The paper includes a wide experimental survey, several ablations, and public code, which is commendable. However, the central mechanistic claim that the learned attention identifies semantically aligned layer pairs is not convincingly validated, and several reporting and baseline weaknesses currently limit the strength of the empirical claims.

major comments (4)
  1. [Allocation with Self-attention, Eq. (3)-(6)] Eq. (3) defines similarity matrices A = R(f)R(f)^T and asserts that the proximity of these Gram matrices measures semantic similarity. No independent evidence supports this proxy; the scale of these matrices depends on feature magnitudes, channel counts, and activation statistics, not just semantic content. Moreover, the attention weights in Eqs. (4)-(5) are trained end-to-end to minimize the same weighted MSE in Eq. (6), so concentrating weight on the easiest-to-match layers is a plausible degenerate solution that would minimize the loss without producing a meaningful semantic correspondence. The STM score in Table 5 is computed from the same Eq. (3) similarity matrices, so it cannot independently validate semantic alignment. I recommend adding an ablation that replaces the learned attention with uniform or fixed attention while keeping the projection layers, and validating the Gram-matrix proxy against an independent semantic similarity measure such as CKA or linear-probe accuracy.
  2. [Experimental Setup / Training Details] The experimental setup states "Results are reported as means (standard deviations) over 3 trials," but Tables 2-7 report only single numbers with no standard deviations. Without these, the significance of small margins (e.g., DVS-CIFAR10 83.60 vs. TET 83.17; CIFAR-10 WRN-28-4 T=2 95.12 vs. KD 94.92) cannot be assessed. Please include the promised standard deviations, or explicitly state that they are omitted and justify why.
  3. [Comparison to Conventional ANN-to-SNN Distillation Methods; Figure 2] The Feature KD baseline is described as the average over different single-layer combination settings. This averaging can severely underestimate the method, especially since Fig. 2 shows that some fixed layer pairs outperform SASTC (e.g., "SNN layer-5 & ANN layer-3" in panel (a)). To establish that SASTC outperforms feature-based distillation, the paper should also report the best single-layer combination result for Feature KD, and ideally implement a multi-layer feature distillation baseline without attention. As written, the comparison is not against a strong feature-KD baseline, and the isolated cases where fixed matching beats SASTC are not fully investigated.
  4. [Abstract; Table 4] The claim that SASTC is the first to make SNNs outperform ANNs on both CIFAR-10 and CIFAR-100 is not supported by a controlled comparison. Table 4 compares the SNN (ResNet-18) to ANN results from Deng et al. (ResNet-19) that are not the teacher models used for distillation (which are ResNet-32x4 for CIFAR-100, etc.), with different architectures and training procedures. Please clarify the reference ANN settings, or perform a same-architecture comparison, and temper or qualify the superlative claim accordingly.
minor comments (7)
  1. [Abstract] The phrase "shedding the new light" should be "shedding new light."
  2. [Eq. (1)] The text says the STM score is the "Average Euclidean Distance" but the formula uses MSE; please define the exact distance used.
  3. [Notations and Background] The statement that the feature pattern is "a permutation rather than a combination" is unclear; a set of layers is normally order-independent. Please clarify what ordering distinction is intended.
  4. [Appendix 1, Figure 1] The figure caption contains garbled artifacts like "/s32/s84/s61/s55"; the figure and caption need to be regenerated cleanly.
  5. [Appendix 2, Table 2] The table header contains a stray fragment "dent) Baseline ANN (teacher)"; the column labels should be cleaned up.
  6. [Figure 2 caption] The caption does not explain the color encoding of the bars or the meaning of the y-axis; please add a legend and axis labels.
  7. [Experiments, Neuromorphic Datasets] The text refers to "robust generalization in noisy-label learning" in connection with neuromorphic datasets, but the noisy-label experiments in Table 7 are on CIFAR-100; please rephrase to avoid suggesting that label noise was introduced to the DVS datasets.

Circularity Check

1 steps flagged · score 2.0 of 10

Self-referential STM diagnostic, but held-out accuracies keep the central claim independent.

  1. self definitional [Eqs. (1), (3)-(5); Section 'SASTC Achieves Semantic Matching during Knowledge Distillation', Table 5]
    "STM score is computed as the Average Euclidean Distance between the generated similarity matrices of each corresponding ANN-SNN feature map pair, as expressed in Equation (1): STMscore = (1/T)(1/|C_t|) sum_t sum_C_t MSE(A_t_sl, A_al). ... Given that feature maps ... tend to cluster at distinct granularities ... the proximity of pairwise similarity matrices serves as a meaningful measure of inherent semantic similarity. These similarity matrices are computed as follows A_t_sl = R(f_t_sl)· R(f_t_sl)' [Eq. (3)]."

    Semantic similarity is defined as Gram-matrix proximity (Eq. 3), and the STM score (Eq. 1) used in Table 5 to conclude 'SASTC Achieves Semantic Matching' is the average MSE of those same Gram matrices. Because the attention weights in Eqs. (4)-(5) are computed from these same A matrices, and the SASTC loss (Eq. 6) is backpropagated through those weights, a lower STM score is partly the training objective's own definition rather than independent evidence of semantic alignment. Still, Eq. (6) is written as feature MSE, not Gram-matrix MSE, and the reported test accuracies are held out, so this is a self-referential diagnostic, not a forced central result.

full rationale

The paper's central quantitative claims (95.12% on CIFAR-10, 79.40% on CIFAR-100 at T=2, 68.69% on ImageNet at T=4, plus neuromorphic results) are held-out test accuracies and are not fitted from the same data; no normalization constant or layer-pair weight is tuned on the test set. The core derivation chain — compute Gram matrices (Eq. 3), derive attention weights (Eqs. 4-5), and use them to weight feature MSE (Eq. 6) — has independent empirical content, and the reported gains over KD, Feature KD, and direct training are external comparisons. The one self-referential element is the use of the STM score (Eq. 1) both as the operational definition of semantic mismatch and as the validation metric (Table 5) for the claim that SASTC achieves semantic matching; since the attention weights are themselves functions of the same A matrices, the STM improvement is a partially circular diagnostic rather than an independent proof of semantic alignment. The paper itself notes in the Fig. 2 discussion that hand-picked fixed layer pairs can outperform SASTC in isolated cases, an honest limitation. Self-citation of Hong et al. 2023 is background related-work only and is not load-bearing; no uniqueness theorem or ansatz is smuggled in via self-citation. The residual concern that Eq. (3)'s Gram-matrix proxy may not faithfully capture semantic similarity is a correctness and validity risk, not a circularity.

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

The central claim rests on several unvalidated modeling choices: the similarity-matrix semantic proxy, the projection networks, the KD loss weight beta, and the fixed SNN hyperparameters. There are no newly invented physical entities. The paper's contribution is an engineering technique, and its results are conditional on these choices.

free parameters (4)
  • beta (loss weight for L_SASTC) = 1300 (CIFAR-10), 700 (CIFAR-100), 50 (ImageNet), 0.01/0.1/1 schedule (neuromorphic)
    Tuned per dataset; the sensitivity analysis shows accuracy varies with beta, so the reported results depend on this choice.
  • temperature alpha for soft targets = 4
    Set for all KD experiments; not swept in the paper, and it affects the KD loss scale.
  • surrogate gradient width gamma = 0.3
    Taken from Deng et al. 2022 and fixed; the student models' training dynamics depend on it.
  • LIF leaky factor lambda and threshold V_th = 0.5 and 1.0
    Fixed neuron parameters shared across all layers; the reported accuracies are conditional on these values.
assumptions (4)
  • domain assumption Pairwise similarity matrices of feature maps are a meaningful measure of semantic similarity between ANN and SNN layers at each time step.
    Stated in 'Allocation with Self-attention' (Eq. 3 and surrounding text); both the attention weights and the STM score depend on this proxy.
  • domain assumption A pretrained ANN provides useful soft targets and intermediate features for supervising an SNN student.
    Standard knowledge-distillation premise; invoked by the KD loss in Eq. (8) and the entire experimental protocol.
  • domain assumption The LIF neuron model with shared leak and threshold together with the triangular surrogate gradient yields trainable gradients.
    Eq. (9)-(11); this is the standard direct-training setup and is not validated against other neuron models in this paper.
  • domain assumption Projection layers can bridge any student layer to any teacher layer's spatial and channel dimensions without degrading semantic content.
    Eq. (2) applies a two-convolution projection to every candidate pair; if the projection destroys semantic information, the attention weights are meaningless.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Self-Attentive Spatio-Temporal Calibration for Precise Intermediate Layer Matching in ANN-to-SNN Distillation." pith.science (2026). https://pith.science/paper/MHP275ZJ

@misc{pith2026250108049,
  author       = {Pith},
  title        = {Pith review of: Self-Attentive Spatio-Temporal Calibration for Precise Intermediate Layer Matching in ANN-to-SNN Distillation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MHP275ZJ}},
  note         = {Machine review of arXiv:2501.08049}
}
read the original abstract

Spiking Neural Networks (SNNs) are promising for low-power computation due to their event-driven mechanism but often suffer from lower accuracy compared to Artificial Neural Networks (ANNs). ANN-to-SNN knowledge distillation can improve SNN performance, but previous methods either focus solely on label information, missing valuable intermediate layer features, or use a layer-wise approach that neglects spatial and temporal semantic inconsistencies, leading to performance degradation.To address these limitations, we propose a novel method called self-attentive spatio-temporal calibration (SASTC). SASTC uses self-attention to identify semantically aligned layer pairs between ANN and SNN, both spatially and temporally. This enables the autonomous transfer of relevant semantic information. Extensive experiments show that SASTC outperforms existing methods, effectively solving the mismatching problem. Superior accuracy results include 95.12% on CIFAR-10, 79.40% on CIFAR-100 with 2 time steps, and 68.69% on ImageNet with 4 time steps for static datasets, and 97.92% on DVS-Gesture and 83.60% on DVS-CIFAR10 for neuromorphic datasets. This marks the first time SNNs have outperformed ANNs on both CIFAR-10 and CIFAR-100, shedding the new light on the potential applications of SNNs.

Figures

Figures reproduced from arXiv: 2501.08049 by the authors.

Figure 1
Figure 1. An overview of the proposed Self-Attentive [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of negative regularization on CIFAR [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Spike Activation Map (SAM) visualization of [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 1
Figure 1. Figure 1: Illustration of temporal dynamics across training [PITH_FULL_IMAGE:figures/full_fig_p010_1.png]
Figure 2
Figure 2. Figure 2: Impact of the hyperparameter 𝛽 [PITH_FULL_IMAGE:figures/full_fig_p011_2.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 16 canonical work pages

  1. [1]

    B.; Manohar, R.; Risk, W

    Akopyan, F.; Sawada, J.; Cassidy, A.; Alvarez-Icaza, R.; Arthur, J.; Merolla, P.; Imam, N.; Nakamura, Y.; Datta, P.; Nam, G.-J.; Taba, B.; Beakes, M.; Brezzo, B.; Kuang, J. B.; Manohar, R.; Risk, W. P.; Jackson, B.; and Modha, D. S. 2015. TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip. IEEE Transactions on Comp...

    work page 2015

  2. [2]

    Bengio, Y.; Courville, A.; and Vincent, P. 2013. Representation Learning: A Review and New Perspectives. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(8): 1798--1828

    work page 2013

  3. [3]

    Bu, T.; Fang, W.; Ding, J.; Dai, P.; Yu, Z.; and Huang, T. 2023. Optimal ANN-SNN conversion for high-accuracy and ultra-low-latency spiking neural networks. arXiv preprint arXiv:2303.04347

    arXiv 2023

  4. [4]

    Cao, Y.; Chen, Y.; and Khosla, D. 2015. Spiking deep convolutional neural networks for energy-efficient object recognition. International Journal of Computer Vision, 113: 54--66

    2015

  5. [5]

    Deng, S.; and Gu, S. 2021. Optimal conversion of conventional artificial neural networks to spiking neural networks. arXiv preprint arXiv:2103.00476

    arXiv 2021

  6. [6]

    Deng, S.; Li, Y.; Zhang, S.; and Gu, S. 2022. Temporal efficient training of spiking neural network via gradient re-weighting. arXiv preprint arXiv:2202.11946

    arXiv 2022

  7. [7]

    U.; Neil, D.; Binas, J.; Cook, M.; Liu, S.-C.; and Pfeiffer, M

    Diehl, P. U.; Neil, D.; Binas, J.; Cook, M.; Liu, S.-C.; and Pfeiffer, M. 2015. Fast-classifying, high-accuracy spiking deep networks through weight and threshold balancing. In 2015 International joint conference on neural networks (IJCNN), 1--8. ieee

    2015

  8. [8]

    Dong, Y.; Zhao, D.; and Zeng, Y. 2024. Temporal Knowledge Sharing Enable Spiking Neural Network Learning From Past and Future . IEEE Transactions on Artificial Intelligence, 5(07): 3524--3534

    work page 2024

Show all 37 references
  1. [9]

    Fang, W.; Yu, Z.; Chen, Y.; Masquelier, T.; Huang, T.; and Tian, Y. 2021. Incorporating learnable membrane time constant to enhance learning of spiking neural networks. In Proceedings of the IEEE/CVF international conference on computer vision, 2661--2671

    2021

  2. [10]

    Han, B.; Srinivasan, G.; and Roy, K. 2020. Rmp-snn: Residual membrane potential neuron for enabling deeper high-accuracy and low-latency spiking neural network. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 13558--13567

    2020

  3. [11]

    Han, D.; Kim, J.; and Kim, J. 2017. Deep pyramidal residual networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, 5927--5935

    2017

  4. [12]

    He, K.; Zhang, X.; Ren, S.; and Sun, J. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 770--778

    2016

  5. [13]

    Hong, D.; Shen, J.; Qi, Y.; and Wang, Y. 2023. LaSNN: Layer-wise ANN-to-SNN Distillation for Effective and Efficient Training in Deep Spiking Neural Networks. arXiv preprint arXiv:2304.09101

    2023 arXiv

  6. [14]

    Hu, Y.; Deng, L.; Wu, Y.; Yao, M.; and Li, G. 2024. Advancing spiking neural networks toward deep residual learning. IEEE Transactions on Neural Networks and Learning Systems

    2024

  7. [15]

    Kim, Y.; and Panda, P. 2021. Visual explanations from spiking neural networks using inter-spike intervals. Scientific reports, 11(1): 1--14

    2021

  8. [16]

    Kugele, A.; Pfeil, T.; Pfeiffer, M.; and Chicca, E. 2020. Efficient processing of spatio-temporal data streams with spiking neural networks. Frontiers in neuroscience, 14: 512192

    2020

  9. [17]

    Lee, D.; Park, S.; Kim, J.; Doh, W.; and Yoon, S. 2021. Energy-efficient Knowledge Distillation for Spiking Neural Networks. CoRR, abs/2106.07172

    2021 arXiv

  10. [18]

    H.; Delbruck, T.; and Pfeiffer, M

    Lee, J. H.; Delbruck, T.; and Pfeiffer, M. 2016. Training deep spiking neural networks using backpropagation. Frontiers in neuroscience, 10: 228000

    2016

  11. [19]

    Maass, W. 1997. Networks of spiking neurons: the third generation of neural network models. Neural networks, 10(9): 1659--1671

    1997

  12. [20]

    A.; Zhao, B.; Serrano, C.; Acha, B.; Serrano-Gotarredona, T.; Chen, S.; and Linares-Barranco, B

    P \'e rez-Carrasco, J. A.; Zhao, B.; Serrano, C.; Acha, B.; Serrano-Gotarredona, T.; Chen, S.; and Linares-Barranco, B. 2013. Mapping from frame-driven to frame-free event-driven vision systems by low-rate rate coding and coincidence processing--application to feedforward Conv...

    2013

  13. [21]

    Rathi, N.; and Roy, K. 2021. 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

    2021

  14. [22]

    Rathi, N.; Srinivasan, G.; Panda, P.; and Roy, K. 2020. Enabling deep spiking neural networks with hybrid conversion and spike timing dependent backpropagation. arXiv preprint arXiv:2005.01807

    2020 arXiv

  15. [23]

    Sengupta, A.; Ye, Y.; Wang, R.; Liu, C.; and Roy, K. 2019. Going Deeper in Spiking Neural Networks: VGG and Residual Architectures. Frontiers in Neuroscience, 13

    2019

  16. [24]

    B.; and Orchard, G

    Shrestha, S. B.; and Orchard, G. 2018. Slayer: Spike layer error reassignment in time. Advances in neural information processing systems, 31

    2018

  17. [25]

    Simonyan, K.; and Zisserman, A. 2014. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556

    2014 arXiv

  18. [26]

    N.; Kaiser, .; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, .; and Polosukhin, I. 2017. Attention is all you need. Advances in neural information processing systems, 30

    2017

  19. [27]

    Wu, Y.; Deng, L.; Li, G.; and Shi, L. 2018. Spatio-temporal backpropagation for training high-performance spiking neural networks. Frontiers in neuroscience, 12: 323875

    2018

  20. [28]

    Wu, Y.; Deng, L.; Li, G.; Zhu, J.; Xie, Y.; and Shi, L. 2019. Direct training for spiking neural networks: Faster, larger, better. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 33, 1311--1318

    2019

  21. [29]

    Wu, Z.; Zhang, H.; Lin, Y.; Li, G.; Wang, M.; and Tang, Y. 2021. Liaf-net: Leaky integrate and analog fire network for lightweight and efficient spatiotemporal information processing. IEEE Transactions on Neural Networks and Learning Systems, 33(11): 6249--6262

    2021

  22. [30]

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

    Xu, Q.; Li, Y.; Shen, J.; Liu, J. K.; Tang, H.; and Pan, G. 2023. Constructing deep spiking neural networks from artificial neural networks with knowledge distillation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 7886--7895

    2023

  23. [31]

    Yao, X.; Li, F.; Mo, Z.; and Cheng, J. 2022. Glif: A unified gated leaky integrate-and-fire neuron for spiking neural networks. Advances in Neural Information Processing Systems, 35: 32160--32171

    2022

  24. [32]

    Zagoruyko, S.; and Komodakis, N. 2016. Wide residual networks. arXiv preprint arXiv:1605.07146

    2016 arXiv

  25. [33]

    Zhang, W.; and Li, P. 2020. Temporal spike sequence learning via backpropagation for deep spiking neural networks. Advances in neural information processing systems, 33: 12022--12033

    2020

  26. [34]

    Zhang, Y.; Liu, X.; Chen, Y.; Peng, W.; Guo, Y.; Huang, X.; and Ma, Z. 2024. Enhancing Representation of Spiking Neural Networks via Similarity-Sensitive Contrastive Learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 16926--16934

    2024

  27. [35]

    Zheng, H.; Wu, Y.; Deng, L.; Hu, Y.; and Li, G. 2021. Going deeper with directly-trained larger spiking neural networks. In Proceedings of the AAAI conference on artificial intelligence, volume 35, 11062--11070

    2021

  28. [36]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  29. [37]

    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 gl...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.