Pith. sign in

REVIEW 4 major objections 4 minor 71 references

Spatio-Temporal Decoupled Learning for Spiking Neural Networks

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Spatio-temporal decoupled learning trains spiking neural networks at BPTT accuracy with 4x lower GPU memory by splitting the network into subnetworks guided by auxiliary networks built from their own subsequent layers.

desk verdict Solid empirical paper on memory-efficient SNN training, with a real but fixable inconsistency in the headline ImageNet memory number. read the letter →

arxiv 2506.01117 v1 pith:7K6JWYGZ submitted 2025-06-01 cs.NE

Chenxiang Ma , Xinyi Chen , Kay Chen Tan , Jibin Wu This is my paper
classification cs.NE MSC 68T07
keywords spikingneuralnetworkslocallearningbackpropagationthroughtimememory-efficienttrainingonlineauxiliarynetworkpartitioningsurrogategradients
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes STDL, a training method for spiking neural networks that splits the network into independently trained subnetworks while still matching the accuracy of full backpropagation through time (BPTT). The payoff is memory: on ImageNet, training SEWResNet-34 with STDL uses roughly 4x less GPU memory than BPTT at comparable accuracy, and the savings grow with the number of time steps. The method works because spatial decoupling is done carefully, using a minimal number of subnetworks under a memory budget and guiding each one with an auxiliary network built from actual subsequent layers, while temporal decoupling drops only gradients whose contribution decays exponentially. The paper argues this is the first local-learning-style SNN training that scales to ImageNet without an accuracy penalty.

What carries the argument

The load-bearing mechanism is the pairing of each subnetwork with an auxiliary network constructed by selecting the largest subset of its subsequent layers under the memory constraint, combined with a greedy partition that minimizes the number of subnetworks. The auxiliary network provides a learned, structure-shaped target so that earlier layers' local losses align their representations with BPTT; the greedy partition keeps subnetworks large enough to preserve synergy. Temporal decoupling then computes gradients online via Eq. (18), using only quantities available at time t and discarding temporally dependent terms whose influence decays like $\lambda^{t'-t}$.

What would settle it

Train STDL but freeze the auxiliary network parameters, or detach the shared downstream layers so earlier subnetworks cannot update them, while keeping everything else identical. If accuracy stays at BPTT level, the overlap is not the source of coupling; if accuracy drops toward ELL's level, the shared-parameter updates are load-bearing.

Watch

Extended reading notes

Core claim

STDL establishes that decoupling both space and time in SNN training need not cost accuracy. Spatial decoupling partitions the network into K subnetworks trained by local losses; a greedy partition minimizes K under a memory constraint, and each subnetwork is paired with an auxiliary network that reuses a largest-capacity subset of its subsequent layers, steering its representations toward those BPTT would produce. Temporal decoupling discards temporally dependent gradients, which decay as $\lambda^{t'-t}$, enabling online updates and making memory independent of the number of time steps. Across seven datasets and several architectures, STDL matches or slightly exceeds BPTT accuracy while cutting GPU memory by 4.0x-4.7x on static image tasks and by up to 20.8x on long-time-step event-based tasks.

Load-bearing premise

The whole accuracy claim rests on the assumption that when an early subnetwork's local loss updates a layer that a later subnetwork also trains, those overlapping updates work together instead of pulling the layer in different directions.

Editorial extensions

If this is right

  • STDL-trained spiking networks can fit on GPUs that cannot hold BPTT's cached states, enabling larger architectures or longer temporal windows at the same memory cost.
  • Because memory use is independent of the number of time steps, STDL is especially favorable for event-based data with high temporal resolution, where BPTT memory grows linearly with time steps.
  • The method transfers across spiking neuron models: on PLIF and ALIF neurons it achieves 14x-24x memory savings while maintaining BPTT-level accuracy.
  • Representation analysis with CKA and linear probing shows STDL layers align closely with BPTT's, suggesting the accuracy parity reflects genuinely similar learned features rather than a lucky loss landscape.
  • The auxiliary networks are used only during training and discarded at inference, so the deployed model is a standard spiking network with no extra cost.

Reading between the lines

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

  • A natural untested ablation would freeze the auxiliary network parameters or detach the shared downstream layers so earlier subnetworks cannot update them; the paper does not isolate whether the coupling benefit comes from representational alignment or from direct multi-loss optimization of overlapping parameters.
  • Since the temporal-gradient cutoff relies on exponential decay with factor $\lambda$, networks with slower membrane decay may need to retain some temporal terms; a threshold based on $\lambda$ rather than omitting all such terms is a testable extension.
  • The partition-plus-auxiliary recipe is not obviously specific to spiking neurons and could be applied to other stateful sequence models where BPTT memory is the bottleneck, provided a decay-like structure justifies dropping long-range temporal gradients.
  • An adaptive memory constraint, loosened as training stabilizes, could further close the small accuracy gap on ImageNet while preserving most of the memory savings.
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 / 4 minor

Summary. The paper proposes spatio-temporal decoupled learning (STDL), a training method for spiking neural networks that partitions a network into subnetworks, trains each subnetwork with an auxiliary network assembled from subsequent layers under a memory budget, and omits temporally dependent gradients to enable online updates. The authors provide a greedy partitioning algorithm with an optimality argument, an information-theoretic motivation for auxiliary network construction, and experiments on seven datasets (CIFAR-10/100, SVHN, ImageNet, CIFAR10-DVS, GAIT-DAY-DVS, HAR-DVS). The headline claims are that STDL matches BPTT accuracy while reducing GPU memory by 4.0x-4.7x, and that it is the first local-learning-style SNN training method to scale to ImageNet at BPTT-level accuracy.

Significance. If the claims hold, this is a significant advance for memory-efficient SNN training: the method combines a local-learning memory profile with accuracy close to BPTT across diverse architectures and datasets, and it ships code and detailed training configurations. The accuracy numbers are internally consistent across Tables II and III, and the ablations in Section V-E support the importance of auxiliary capacity. However, the headline memory ratio rests on an uncontrolled batch-size comparison, the central theoretical proposition is not actually proved, and the ImageNet comparison lacks error bars; these issues must be resolved before the stated contributions can be accepted.

major comments (4)
  1. [Table II, Appendix B, and Table VII] The image-classification table states in its caption that all GPU memory is measured at batch size 512, but Appendix B reports that BPTT on ImageNet was run at batch size 256 because of memory constraints, and Table VII also distinguishes the ImageNet finetune batch size from the baseline batch size. The reported BPTT memory of 424.35 GB therefore corresponds to batch 256, while the STDL memory of 106.53 GB corresponds to batch 512. Since activation memory scales roughly linearly with batch size, the 4.0x ratio cited in the abstract and Section V-A2 is not computed at a fixed operating point; at a common batch size the true ratio would be approximately 8x (at batch 512) or 2x (at batch 256). The authors should re-measure at a fixed batch size or clearly report the corrected factor, because the exact memory saving is a central claim of the paper.
  2. [Section IV-C, Proposition 2] Proposition 2 is not a valid proof. The construction of a function-preserving expanded network via zero-padding and identity residual connections is not justified for the actual auxiliary networks, which include spiking activations, downsampling layers, and normalization; moreover, the step claiming that an expanded network has 'the potential to discover' a more informative representation is an assertion rather than a derivation. Since this proposition is used to motivate the depth/width maximization, the authors should either replace it with a theorem under explicit assumptions or clearly present the capacity-growing rule as a heuristic that is empirically supported by the ablations in Section V-E.
  3. [Sections IV-A, IV-C, and V-E] The auxiliary network for each subnetwork is built by reusing actual subsequent layers (including the classifier), and those same layers are later trained as part of later subnetworks. The stop-gradient operator in Eq. (10) blocks gradient flow from later subnetworks into the current subnetwork's representation, but it does not prevent the local loss of subnetwork k from updating the parameters of its auxiliary network G^k, which are shared with later subnetworks. The paper does not analyze or ablate this overlapping-gradient coupling. An explicit experiment, such as training the auxiliary network with a detached copy or frozen weights, or measuring gradient conflict between the overlapping losses, is needed to support the claim that the reuse is beneficial rather than harmful.
  4. [Section V-A, Table II; Section V-B, Table IV] The central claim of BPTT-comparable accuracy on ImageNet rests on single runs (BPTT 70.12 vs STDL 69.87, with no standard deviation), and the HAR-DVS rows in Table IV also lack error bars. Given that the other datasets show seed-to-seed variation on the order of 0.1-0.2 percentage points, a single ImageNet run is insufficient evidence to support the phrase 'comparable accuracy to BPTT' on large-scale data. The authors should report multiple seeds, or at minimum provide a clear statistical justification for the single-run comparison.
minor comments (4)
  1. [Figure 9] The y-axis labels and legend in Figure 9 are rendered as unreadable encoded text in the manuscript; please ensure the figure is regenerated with correct fonts.
  2. [Section IV-D] The statement that temporally dependent gradients 'contribute negligibly' is supported only by an exponential-decay argument and a citation to prior work; reporting an empirical measurement of the omitted-gradient norm would make the approximation more convincing.
  3. [Table III] The comparison with published BPTT-based methods uses results from their original papers under different training settings (e.g., different numbers of time steps); a sentence noting that this is an informal comparison would help readers interpret the table.
  4. [Table VIII] The symbol rho is defined as a memory efficiency ratio but is never used elsewhere in the text; either explain its role or remove it to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity is found; the central accuracy claim is benchmarked against external BPTT baselines, and the ImageNet memory-ratio discrepancy is a measurement-consistency concern rather than a circular step.

full rationale

STDL's derivation chain is self-contained against external benchmarks. The accuracy claims are produced by training with local losses (Eqs. 10, 11, 17, 18) on standard datasets and compared with independently re-implemented BPTT, SLTT, DECOLLE, and ELL baselines; no parameter is fitted to a subset of the reported results and then presented as a prediction. The auxiliary-network construction reuses subsequent layers to maximize capacity under a memory budget, and although it is motivated by alignment with BPTT, BPTT representations are not used as training targets, so the later CKA and linear-probing similarity analyses are post-hoc evaluations rather than alignment-by-construction results. The greedy partitioning optimality is proven in the paper (Proposition 1), and the mutual-information argument (Proposition 2) is a heuristic monotonicity claim rather than a fitted input. Self-citations [18] and [19] provide context on supervised local learning but do not carry the central result. Appendix B does reveal that BPTT on ImageNet was run at batch size 256 while STDL used batch size 512, which makes the exact 4.0x memory ratio in Table II a measurement-protocol consistency issue; however, this is a correctness concern, not circularity. No step in the paper reduces by definition or by a self-citation chain to its own inputs.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

No new physical or conceptual entities are introduced; subnetworks and auxiliary networks are configurations of existing layers. The main uncharged inputs are the hand-chosen memory constraint, the neuronal decay hyperparameter, the negligibility of temporal gradients, and the informal capacity monotonicity assumption in Proposition 2.

free parameters (2)
  • memory constraint per subnetwork M_sub = rho = 0.6 to 0.7 (Table VIII)
    Chosen by hand for each architecture; determines the number of subnetworks and the size of auxiliary networks, and directly controls the accuracy versus memory trade-off shown in Fig. 10.
  • neuronal decay factor lambda = 0.1 (Table VII)
    The temporal gradient omission is justified by the exponential decay lambda^(t'-t). Accuracy of the method depends on this hyperparameter, which is set to the same value across datasets.
assumptions (3)
  • domain assumption Temporally dependent gradients contribute negligibly to parameter updates.
    STDL discards all temporally dependent gradients to enable online learning (Eq. 16 and surrounding text). The paper cites prior SLTT results rather than proving this holds for its subnetwork and auxiliary setup.
  • ad hoc to paper Larger auxiliary networks, when sufficiently trained, attain at least as much label-related mutual information.
    Proposition 2 assumes the expanded network can be initialized to copy the smaller one and then has the potential to discover more informative representations, which is not a guarantee and is not formally established.
  • domain assumption The Markov chain y -> x -> s^k -> yhat holds for the DPI-based lower bound.
    Reasonable when s^k is a deterministic function of x, but the paper treats representations as random variables without formalizing the generative model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spatio-Temporal Decoupled Learning for Spiking Neural Networks." pith.science (2026). https://pith.science/paper/7K6JWYGZ

@misc{pith2026250601117,
  author       = {Pith},
  title        = {Pith review of: Spatio-Temporal Decoupled Learning for Spiking Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7K6JWYGZ}},
  note         = {Machine review of arXiv:2506.01117}
}
read the original abstract

Spiking neural networks (SNNs) have gained significant attention for their potential to enable energy-efficient artificial intelligence. However, effective and efficient training of SNNs remains an unresolved challenge. While backpropagation through time (BPTT) achieves high accuracy, it incurs substantial memory overhead. In contrast, biologically plausible local learning methods are more memory-efficient but struggle to match the accuracy of BPTT. To bridge this gap, we propose spatio-temporal decouple learning (STDL), a novel training framework that decouples the spatial and temporal dependencies to achieve both high accuracy and training efficiency for SNNs. Specifically, to achieve spatial decoupling, STDL partitions the network into smaller subnetworks, each of which is trained independently using an auxiliary network. To address the decreased synergy among subnetworks resulting from spatial decoupling, STDL constructs each subnetwork's auxiliary network by selecting the largest subset of layers from its subsequent network layers under a memory constraint. Furthermore, STDL decouples dependencies across time steps to enable efficient online learning. Extensive evaluations on seven static and event-based vision datasets demonstrate that STDL consistently outperforms local learning methods and achieves comparable accuracy to the BPTT method with considerably reduced GPU memory cost. Notably, STDL achieves 4x reduced GPU memory than BPTT on the ImageNet dataset. Therefore, this work opens up a promising avenue for memory-efficient SNN training. Code is available at https://github.com/ChenxiangMA/STDL.

Figures

Figures reproduced from arXiv: 2506.01117 by the authors.

Figure 1
Figure 1. Comparison of BPTT, supervised local learning rules (DECOLLE [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Comparison of BPTT, supervised local learning, and our STDL for training SNNs. (a) BPTT caches the entire trajectory of neuronal states during forward propagation for gradient computation, with memory overhead scaled along with both the number of layers and time steps. (b) Supervised local learning caches the states of only one layer at each time step, as it enables the online learning of each network layer along wi… view at source ↗
Figure 3
Figure 3. Comparison of layer-wise linear separability among supervised local [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Illustration of auxiliary network construction in STDL. For each subnetwork, the auxiliary network is constructed by selecting a subset of its subsequent layers. The subnet has the largest expressive capacity, i.e., depth and width, while not exceeding a given memory c…
Figure 5
Figure 5. Figure 5: Visualization of event streams (accumulated over 10ms) in CIFAR10- DVS [34], GAIT-DAY-DVS [35], and HAR-DVS [36] from top to bottom. 0 5 10 15 20 25 30 35 40 Depth-Time Index 0 2 4 6 8 10 12 14 16 GPU Memory (GB) BPTT STDL [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: GPU memory usage pattern with respect to the layer and time index. ResNet-18 is trained with BPTT and STDL on CIFAR-10 with 4 steps. up. Notably, the non-linear increase in GPU memory usage as the network depth grows reflects the decreasing memory footprint of the laye…
Figure 7
Figure 7. Figure 7: Comparison of layer-wise representation similarity. We utilize CKA [71] to measure the layer-wise similarity of representations between BPTT and other learning rules. To provide a fair baseline for BPTT, we measure the similarity between two networks trained with diffe…
Figure 8
Figure 8. Figure 8: Comparison of layer-wise linear separability among STDL and baselines. freeze the parameters of well-trained networks and further train additional linear classifiers that are attached to each hidden layer. The results are provided in [PITH_FULL_IMAGE:figures/full_fig_…
Figure 10
Figure 10. Figure 10: Trade-offs between GPU memory efficiency and accuracy in STDL. The dashed line denotes BPTT’s accuracy. GPU memory ratio is the memory of BPTT divided by that of STDL. TABLE VI GENERALIZATION OF STDL ON DIFFERENT SPIKING NEURON MODELS. THE GAIT-DAY-DVS DATASET IS ADOP…
Figure 9
Figure 9. Figure 9: Influences of auxiliary networks and subnetworks. Each row denotes a variant of STDL with a modification described on the y-axis. The blue bar chart reports each variant’s accuracy for ResNet-18 on CIFAR-10. TABLE V COMPARISON OF ACCURACY AND GPU MEMORY USAGE BETWEEN S…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

71 extracted references · 40 canonical work pages

  1. [1]

    Dayan and L

    P. Dayan and L. F. Abbott,Theoretical neuroscience: computational and mathematical modeling of neural systems. Cambridge, MA, USA: MIT press, 2001, vol. 806

    work page 2001

  2. [2]

    Evolution, brain, and the nature of language,

    R. C. Berwick, A. D. Friederici, N. Chomsky, and J. J. Bolhuis, “Evolution, brain, and the nature of language,”Trends in Cognitive Sciences, vol. 17, no. 2, pp. 89–98, 2013

    work page 2013

  3. [3]

    A deep learning framework for neuroscience,

    B. A. Richards, T. P. Lillicrap, P. Beaudoin, Y . Bengio, R. Bogacz, A. Christensen, C. Clopath, R. P. Costa, A. de Berker, S. Ganguliet al., “A deep learning framework for neuroscience,”Nature Neuroscience, vol. 22, no. 11, pp. 1761–1770, 2019

    work page 2019

  4. [4]

    Cat- alyzing next-generation artificial intelligence through neuroai,

    A. Zador, S. Escola, B. Richards, B. ¨Olveczky, Y . Bengio, K. Boahen, M. Botvinick, D. Chklovskii, A. Churchland, C. Clopathet al., “Cat- alyzing next-generation artificial intelligence through neuroai,”Nature Communications, vol. 14, no. 1, p. 1597, 2023

    work page 2023

  5. [5]

    Networks of spiking neurons: the third generation of neural network models,

    W. Maass, “Networks of spiking neurons: the third generation of neural network models,”Neural Networks, vol. 10, no. 9, pp. 1659–1671, 1997

    1997

  6. [6]

    Towards spike-based machine intelligence with neuromorphic computing,

    K. Roy, A. Jaiswal, and P. Panda, “Towards spike-based machine intelligence with neuromorphic computing,”Nature, vol. 575, no. 7784, pp. 607–617, 2019

    2019

  7. [7]

    Deep learning,

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning,”Nature, vol. 521, no. 7553, pp. 436–444, 2015

    2015

  8. [8]

    Toward ultralow- power neuromorphic speech enhancement with spiking-fullsubnet,

    X. Hao, C. Ma, Q. Yang, J. Wu, and K. C. Tan, “Toward ultralow- power neuromorphic speech enhancement with spiking-fullsubnet,” IEEE Transactions on Neural Networks and Learning Systems, pp. 1–15, 2025

    work page 2025

Show all 71 references
  1. [9]

    Loihi: A neuromorphic manycore processor with on-chip learning,

    M. Davies, N. Srinivasa, T.-H. Lin, G. Chinya, Y . Cao, S. H. Choday, G. Dimou, P. Joshi, N. Imam, S. Jainet al., “Loihi: A neuromorphic manycore processor with on-chip learning,”IEEE Micro, vol. 38, no. 1, pp. 82–99, 2018

    2018

  2. [10]

    Towards artificial general intelligence with hybrid tianjic chip architecture,

    J. Pei, L. Deng, S. Song, M. Zhao, Y . Zhang, S. Wu, G. Wang, Z. Zou, Z. Wu, W. Heet al., “Towards artificial general intelligence with hybrid tianjic chip architecture,”Nature, vol. 572, no. 7767, pp. 106–111, 2019

    2019

  3. [11]

    Brain- inspired computing: A systematic survey and future trends,

    G. Li, L. Deng, H. Tang, G. Pan, Y . Tian, K. Roy, and W. Maass, “Brain- inspired computing: A systematic survey and future trends,”Proceedings of the IEEE, vol. 112, no. 6, pp. 544–584, 2024

    2024

  4. [12]

    Training spiking neural networks using lessons from deep learning,

    J. K. Eshraghian, M. Ward, E. O. Neftci, X. Wang, G. Lenz, G. Dwivedi, M. Bennamoun, D. S. Jeong, and W. D. Lu, “Training spiking neural networks using lessons from deep learning,”Proceedings of the IEEE, 2023

    2023

  5. [13]

    Backpropagation through time: what it does and how to do it,

    P. J. Werbos, “Backpropagation through time: what it does and how to do it,”Proceedings of the IEEE, vol. 78, no. 10, pp. 1550–1560, 1990

    1990

  6. [14]

    Spatio-temporal backpropa- gation for training high-performance spiking neural networks,

    Y . Wu, L. Deng, G. Li, J. Zhu, and L. Shi, “Spatio-temporal backpropa- gation for training high-performance spiking neural networks,”Frontiers in Neuroscience, vol. 12, p. 331, 2018

    2018

  7. [15]

    Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based opti- mization to spiking neural networks,

    E. O. Neftci, H. Mostafa, and F. Zenke, “Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based opti- mization to spiking neural networks,”IEEE Signal Processing Magazine, vol. 36, no. 6, pp. 51–63, 2019

    2019

  8. [16]

    Learnable surrogate gradient for direct training spiking neural networks,

    S. Lian, J. Shen, Q. Liu, Z. Wang, R. Yan, and H. Tang, “Learnable surrogate gradient for direct training spiking neural networks,” in Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, 2023, pp. 3002–3010

    2023

  9. [17]

    Brain-inspired learning on neuromorphic substrates,

    F. Zenke and E. O. Neftci, “Brain-inspired learning on neuromorphic substrates,”Proceedings of the IEEE, vol. 109, no. 5, pp. 935–950, 2021

    2021

  10. [18]

    Deep spike learning with local classifiers,

    C. Ma, R. Yan, Z. Yu, and Q. Yu, “Deep spike learning with local classifiers,”IEEE Transactions on Cybernetics, vol. 53, no. 5, pp. 3363– 3375, 2023

    2023

  11. [19]

    Scaling supervised local learning with augmented auxiliary networks,

    C. Ma, J. Wu, C. Si, and K. Tan, “Scaling supervised local learning with augmented auxiliary networks,” inThe Twelfth International Conference on Learning Representations, 2024

    2024

  12. [20]

    Synaptic modifications in cultured hip- pocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type,

    G.-q. Bi and M.-m. Poo, “Synaptic modifications in cultured hip- pocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type,”Journal of Neuroscience, vol. 18, no. 24, pp. 10 464–10 472, 1998

    1998

  13. [21]

    Unsupervised learning of digit recognition using spike-timing-dependent plasticity,

    P. U. Diehl and M. Cook, “Unsupervised learning of digit recognition using spike-timing-dependent plasticity,”Frontiers in Computational Neuroscience, vol. 9, p. 99, 2015

    2015

  14. [22]

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

    S. R. Kheradpisheh, M. Ganjtabesh, S. J. Thorpe, and T. Masquelier, “STDP-based spiking deep convolutional neural networks for object recognition,”Neural Networks, vol. 99, pp. 56–67, 2018

    2018

  15. [23]

    A biologically plausible supervised learning method for spiking neural networks using the symmetric stdp rule,

    Y . Hao, X. Huang, M. Dong, and B. Xu, “A biologically plausible supervised learning method for spiking neural networks using the symmetric stdp rule,”Neural Networks, vol. 121, pp. 387–395, 2020

    2020

  16. [24]

    Online supervised learning for hardware- based multilayer spiking neural networks through the modulation of weight-dependent spike-timing-dependent plasticity,

    N. Zheng and P. Mazumder, “Online supervised learning for hardware- based multilayer spiking neural networks through the modulation of weight-dependent spike-timing-dependent plasticity,”IEEE Transactions on Neural Networks and Learning Systems, vol. 29, no. 9, pp. 4287– 4302, 2018

    2018

  17. [25]

    Memory-dependent computation and learning in spiking neural networks through hebbian plasticity,

    T. Limbacher, O. ¨Ozdenizci, and R. Legenstein, “Memory-dependent computation and learning in spiking neural networks through hebbian plasticity,”IEEE Transactions on Neural Networks and Learning Sys- tems, vol. 36, no. 2, pp. 2551–2562, 2025

    2025

  18. [26]

    S-TLLR: STDP-inspired temporal local learning rule for spiking neural networks,

    M. P. E. Apolinario and K. Roy, “S-TLLR: STDP-inspired temporal local learning rule for spiking neural networks,”Transactions on Machine Learning Research, 2025

    2025

  19. [27]

    Synaptic plasticity dynamics for deep continuous local learning (DECOLLE),

    J. Kaiser, H. Mostafa, and E. Neftci, “Synaptic plasticity dynamics for deep continuous local learning (DECOLLE),”Frontiers in Neuroscience, vol. 14, p. 424, 2020

    2020

  20. [28]

    Deep residual learning in spiking neural networks,

    W. Fang, Z. Yu, Y . Chen, T. Huang, T. Masquelier, and Y . Tian, “Deep residual learning in spiking neural networks,”Advances in Neural Information Processing Systems, vol. 34, pp. 21 056–21 069, 2021

    2021

  21. [29]

    Very deep convolutional networks for large-scale image recognition,

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

    2014 arXiv

  22. [30]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 770–778

    2016

  23. [31]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, G. Hintonet al., “Learning multiple layers of features from tiny images,”Master’s thesis, Department of Computer Science, University of Toronto, 2009

    2009

  24. [32]

    Reading digits in natural images with unsupervised feature learning,

    Y . Netzer, T. Wang, A. Coates, A. Bissacco, B. Wu, A. Y . Nget al., “Reading digits in natural images with unsupervised feature learning,” in NIPS Workshop on Deep Learning and Unsupervised Feature Learning, vol. 2011, no. 5. Granada, Spain, 2011, p. 7

    2011

  25. [33]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in2009 IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2009, pp. 248–255

    2009

  26. [34]

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

    H. Li, H. Liu, X. Ji, G. Li, and L. Shi, “Cifar10-dvs: an event-stream dataset for object classification,”Frontiers in Neuroscience, vol. 11, p. 244131, 2017

    2017

  27. [35]

    Event-stream representation for human gaits identification using deep neural networks,

    Y . Wang, X. Zhang, Y . Shen, B. Du, G. Zhao, L. Cui, and H. Wen, “Event-stream representation for human gaits identification using deep neural networks,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 7, pp. 3436–3449, 2022

    2022

  28. [36]

    Hardvs: Revisiting human activity recognition with dynamic vision sensors,

    X. Wang, Z. Wu, B. Jiang, Z. Bao, L. Zhu, G. Li, Y . Wang, and Y . Tian, “Hardvs: Revisiting human activity recognition with dynamic vision sensors,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 6, pp. 5615–5623, Mar. 2024. 14

    2024

  29. [37]

    Differen- tiable spike: Rethinking gradient-descent for training spiking neural networks,

    Y . Li, Y . Guo, S. Zhang, S. Deng, Y . Hai, and S. Gu, “Differen- tiable spike: Rethinking gradient-descent for training spiking neural networks,”Advances in Neural Information Processing Systems, vol. 34, pp. 23 426–23 439, 2021

    2021

  30. [38]

    Adaptive smoothing gradient learning for spiking neural networks,

    Z. Wang, R. Jiang, S. Lian, R. Yan, and H. Tang, “Adaptive smoothing gradient learning for spiking neural networks,” inInternational Confer- ence on Machine Learning. PMLR, 2023, pp. 35 798–35 816

    2023

  31. [39]

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

    S. Deng, Y . Li, S. Zhang, and S. Gu, “Temporal efficient training of spiking neural network via gradient re-weighting,” inInternational Conference on Learning Representations, 2021

    2021

  32. [40]

    Recdis-snn: Rectifying membrane potential distribution for directly training spiking neural networks,

    Y . Guo, X. Tong, Y . Chen, L. Zhang, X. Liu, Z. Ma, and X. Huang, “Recdis-snn: Rectifying membrane potential distribution for directly training spiking neural networks,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 326– 335

    2022

  33. [41]

    Going deeper with directly-trained larger spiking neural networks,

    H. Zheng, Y . Wu, L. Deng, Y . Hu, and G. Li, “Going deeper with directly-trained larger spiking neural networks,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 12, 2021, pp. 11 062–11 070

    2021

  34. [42]

    Temporal effective batch normalization in spiking neural networks,

    C. Duan, J. Ding, S. Chen, Z. Yu, and T. Huang, “Temporal effective batch normalization in spiking neural networks,”Advances in Neural Information Processing Systems, vol. 35, pp. 34 377–34 390, 2022

    2022

  35. [43]

    Incorporating learnable membrane time constant to enhance learning of spiking neural networks,

    W. Fang, Z. Yu, Y . Chen, T. Masquelier, T. Huang, and Y . Tian, “Incorporating learnable membrane time constant to enhance learning of spiking neural networks,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 2661–2671

    2021

  36. [44]

    Accurate and efficient time-domain classification with adaptive spiking recurrent neural networks,

    B. Yin, F. Corradi, and S. M. Boht ´e, “Accurate and efficient time-domain classification with adaptive spiking recurrent neural networks,”Nature Machine Intelligence, vol. 3, no. 10, pp. 905–913, 2021

    2021

  37. [45]

    Glif: A unified gated leaky integrate-and-fire neuron for spiking neural networks,

    X. Yao, F. Li, Z. Mo, and J. Cheng, “Glif: A unified gated leaky integrate-and-fire neuron for spiking neural networks,”Advances in Neural Information Processing Systems, vol. 35, pp. 32 160–32 171, 2022

    2022

  38. [46]

    Attention spiking neural networks,

    M. Yao, G. Zhao, H. Zhang, Y . Hu, L. Deng, Y . Tian, B. Xu, and G. Li, “Attention spiking neural networks,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 8, pp. 9393–9410, 2023

    2023

  39. [47]

    Attention-based deep spiking neural networks for temporal credit assignment problems,

    L. Qin, Z. Wang, R. Yan, and H. Tang, “Attention-based deep spiking neural networks for temporal credit assignment problems,”IEEE Trans- actions on Neural Networks and Learning Systems, vol. 35, no. 8, pp. 10 301–10 311, 2024

    2024

  40. [48]

    A hybrid neural coding approach for pattern recognition with spiking neural networks,

    X. Chen, Q. Yang, J. Wu, H. Li, and K. C. Tan, “A hybrid neural coding approach for pattern recognition with spiking neural networks,”IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1–15, 2023

    2023

  41. [49]

    Synaptic learning with augmented spikes,

    Q. Yu, S. Song, C. Ma, L. Pan, and K. C. Tan, “Synaptic learning with augmented spikes,”IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 3, pp. 1134–1146, 2022

    2022

  42. [50]

    Construct- ing accurate and efficient deep spiking neural networks with double- threshold and augmented schemes,

    Q. Yu, C. Ma, S. Song, G. Zhang, J. Dang, and K. C. Tan, “Construct- ing accurate and efficient deep spiking neural networks with double- threshold and augmented schemes,”IEEE Transactions on Neural Net- works and Learning Systems, vol. 33, no. 4, pp. 1714–1726, 2022

    2022

  43. [51]

    Tc-lif: A two- compartment spiking neuron model for long-term sequential modelling,

    S. Zhang, Q. Yang, C. Ma, J. Wu, H. Li, and K. C. Tan, “Tc-lif: A two- compartment spiking neuron model for long-term sequential modelling,” inProceedings of the AAAI conference on artificial intelligence, vol. 38, no. 15, 2024, pp. 16 838–16 847

    2024

  44. [52]

    A parallel multi- compartment spiking neuron for multi-scale sequential modeling,

    X. Chen, J. Wu, C. Ma, Y . Yan, and K. Tan, “A parallel multi- compartment spiking neuron for multi-scale sequential modeling,” 2024

    2024

  45. [53]

    Brain-inspired machine intelligence: A survey of neurobiologically-plausible credit assignment,

    A. G. Ororbia, “Brain-inspired machine intelligence: A survey of neurobiologically-plausible credit assignment,”arXiv preprint arXiv:2312.09257, 2023

    2023 arXiv

  46. [54]

    Training spiking neural networks with local tandem learning,

    Q. Yang, J. Wu, M. Zhang, Y . Chua, X. Wang, and H. Li, “Training spiking neural networks with local tandem learning,” inAdvances in Neural Information Processing Systems, S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, Eds., vol. 35. Curran Associates, Inc....

    2022

  47. [55]

    Brain-inspired global-local learning incorpo- rated with neuromorphic computing,

    Y . Wu, R. Zhao, J. Zhu, F. Chen, M. Xu, G. Li, S. Song, L. Deng, G. Wang, H. Zhenget al., “Brain-inspired global-local learning incorpo- rated with neuromorphic computing,”Nature Communications, vol. 13, no. 1, p. 65, 2022

    2022

  48. [56]

    Stsf: Spiking time sparse feedback learning for spiking neural networks,

    P. He, R. Xiao, C. Tang, S. Huang, J. Lv, and H. Tang, “Stsf: Spiking time sparse feedback learning for spiking neural networks,”IEEE Transactions on Neural Networks and Learning Systems, pp. 1–14, 2025

    2025

  49. [57]

    A solution to the learning dilemma for recurrent networks of spiking neurons,

    G. Bellec, F. Scherr, A. Subramoney, E. Hajek, D. Salaj, R. Legenstein, and W. Maass, “A solution to the learning dilemma for recurrent networks of spiking neurons,”Nature Communications, vol. 11, p. 3625, 2020

    2020

  50. [58]

    Online training through time for spiking neural networks,

    M. Xiao, Q. Meng, Z. Zhang, D. He, and Z. Lin, “Online training through time for spiking neural networks,”Advances in Neural Infor- mation Processing Systems, vol. 35, pp. 20 717–20 730, 2022

    2022

  51. [59]

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

    Q. Meng, M. Xiao, S. Yan, Y . Wang, Z. Lin, and Z.-Q. Luo, “Towards memory-and time-efficient backpropagation for training spiking neural networks,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 6166–6176

    2023

  52. [60]

    Accurate online training of dynamical spiking neural networks through forward propagation through time,

    B. Yin, F. Corradi, and S. M. Boht ´e, “Accurate online training of dynamical spiking neural networks through forward propagation through time,”Nature Machine Intelligence, vol. 5, no. 5, pp. 518–527, 2023

    2023

  53. [61]

    Spiking deep convolutional neural networks for energy-efficient object recognition,

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

    2015

  54. [62]

    Toward high-accuracy and low-latency spiking neural networks with two-stage optimization,

    Z. Wang, Y . Zhang, S. Lian, X. Cui, R. Yan, and H. Tang, “Toward high-accuracy and low-latency spiking neural networks with two-stage optimization,”IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 2, pp. 3189–3203, 2025

    2025

  55. [63]

    A tandem learning rule for effective training and rapid inference of deep spiking neural networks,

    J. Wu, Y . Chua, M. Zhang, G. Li, H. Li, and K. C. Tan, “A tandem learning rule for effective training and rapid inference of deep spiking neural networks,”IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 1, pp. 446–460, 2023

    2023

  56. [64]

    Progressive tandem learning for pattern recognition with deep spiking neural networks,

    J. Wu, C. Xu, X. Han, D. Zhou, M. Zhang, H. Li, and K. C. Tan, “Progressive tandem learning for pattern recognition with deep spiking neural networks,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 11, pp. 7824–7840, 2022

    2022

  57. [65]

    The organization of behavior: a neuropsycholocigal theory,

    D. O. Hebb, “The organization of behavior: a neuropsycholocigal theory,”A Wiley Book in Clinical Psychology, vol. 62, p. 78, 1949

    1949

  58. [66]

    Opening the black box of deep neural networks via information,

    R. Shwartz-Ziv and N. Tishby, “Opening the black box of deep neural networks via information,”arXiv preprint arXiv:1703.00810, 2017

    2017 arXiv

  59. [67]

    The power of depth for feedforward neural networks,

    R. Eldan and O. Shamir, “The power of depth for feedforward neural networks,” inConference on Learning Theory. PMLR, 2016, pp. 907– 940

    2016

  60. [68]

    IM-loss: Information maximization loss for spiking neural networks,

    Y . Guo, Y . Chen, L. Zhang, X. Liu, Y . Wang, X. Huang, and Z. Ma, “IM-loss: Information maximization loss for spiking neural networks,” inAdvances in Neural Information Processing Systems, A. H. Oh, A. Agarwal, D. Belgrave, and K. Cho, Eds., 2022

    2022

  61. [69]

    Spikingjelly: An open-source machine learning infrastructure platform for spike-based intelligence,

    W. Fang, Y . Chen, J. Ding, Z. Yu, T. Masquelier, D. Chen, L. Huang, H. Zhou, G. Li, and Y . Tian, “Spikingjelly: An open-source machine learning infrastructure platform for spike-based intelligence,”Science Advances, vol. 9, no. 40, p. eadi1480, 2023

    2023

  62. [70]

    Inherent redundancy in spiking neural networks,

    M. Yao, J. Hu, G. Zhao, Y . Wang, Z. Zhang, B. Xu, and G. Li, “Inherent redundancy in spiking neural networks,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2023, pp. 16 924–16 934

    2023

  63. [71]

    Similarity of neural network representations revisited,

    S. Kornblith, M. Norouzi, H. Lee, and G. Hinton, “Similarity of neural network representations revisited,” inProceedings of the 36th International Conference on Machine Learning, ser. Proceedings of Machine Learning Research, K. Chaudhuri and R. Salakhutdinov, Eds., vol. 97. P...

    2019

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.