pith. sign in

arxiv: 2505.05375 · v3 · submitted 2025-05-08 · 💻 cs.CV · cs.AI· cs.LG· cs.NE

Threshold Modulation for Online Test-Time Adaptation of Spiking Neural Networks

Kejie Zhao , Wenjia Hua , Aiersi Tuerhong , Luziwei Leng , Yuxin Ma , Qinghai Guo This is my paper

Pith reviewed 2026-05-22 15:55 UTC · model grok-4.3

classification 💻 cs.CV cs.AIcs.LGcs.NE
keywords spiking neural networksonline test-time adaptationthreshold modulationneuromorphic hardwaredistribution shiftsedge computingadaptation methods
0
0 comments X

The pith

Spiking neural networks adapt to distribution shifts by dynamically modulating their firing thresholds at test time.

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

The paper proposes Threshold Modulation as an online test-time adaptation method specifically for spiking neural networks. It works by adjusting neuron firing thresholds using a normalization inspired by biological neuronal dynamics. This addresses the challenge of SNNs losing performance when input data distributions change after the model is deployed on edge devices. Existing adaptation techniques for standard neural networks are not suitable for the event-driven, low-power nature of SNNs and neuromorphic hardware. If successful, this would allow SNNs to maintain accuracy in real-world scenarios with changing conditions without needing original training data or extra labels.

Core claim

The central discovery is that Threshold Modulation (TM), by dynamically adjusting the firing threshold through neuronal dynamics-inspired normalization, enables effective online test-time adaptation for SNNs. This improves robustness against distribution shifts while keeping computational cost low and ensuring compatibility with neuromorphic chips.

What carries the argument

Threshold Modulation (TM) that dynamically adjusts the firing threshold of neurons using normalization inspired by neuronal dynamics to counteract distribution shifts in an online manner.

If this is right

  • Models deployed on neuromorphic hardware can adapt to new environments without access to source data.
  • Adaptation preserves the low-power advantages of SNNs.
  • Performance improves on benchmark datasets with distribution shifts compared to non-adapted SNNs.
  • The method provides a practical framework that can inspire designs for future neuromorphic chips.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This could lead to SNNs being used in more dynamic real-world applications like autonomous systems where data changes over time.
  • Similar threshold adjustment ideas might apply to other spiking-based models or even non-spiking ones for efficiency.
  • Integration with other low-power techniques could further enhance edge device performance under varying conditions.

Load-bearing premise

Dynamically adjusting the firing threshold through neuronal dynamics-inspired normalization is enough to reduce the effects of distribution shifts on SNN accuracy while staying compatible with hardware limits.

What would settle it

Running the method on standard benchmarks with induced distribution shifts and finding no significant accuracy gain over a fixed-threshold baseline or seeing increased power consumption.

Figures

Figures reproduced from arXiv: 2505.05375 by Aiersi Tuerhong, Kejie Zhao, Luziwei Leng, Qinghai Guo, Wenjia Hua, Yuxin Ma.

Figure 1
Figure 1. Figure 1: Framework overview. (a) In the pre-training phase, membrane potentials are normalized after neuronal charging and affine parameters are trained. Before test-time adaptation, the model can be mapped and deployed on a neuromorphic chip after threshold re-parameterization. (b) During adaptation on-chip, statistics are updated to modulate the firing threshold. The affine parameters can also be optimized upon n… view at source ↗
Figure 2
Figure 2. Figure 2: Firing rate visualization on CIFAR-10-C with Contrast. Upper: Density Estimation of firing rates of the last convolutional layer in Spiking ResNet￾20. Lower: Histogram of firing rates of the two layers in Spiking VGG-16m. The firing rate distribution was shifted away by corrupted data (’Source’); TM-NORM and TM-ENT bring it closer to the clean data. the theoretical energy consumption of inference under the… view at source ↗
Figure 3
Figure 3. Figure 3: Test entropy, running accuracy and the value of affine parameters as [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
read the original abstract

Recently, spiking neural networks (SNNs), deployed on neuromorphic chips, provide highly efficient solutions on edge devices in different scenarios. However, their ability to adapt to distribution shifts after deployment has become a crucial challenge. Online test-time adaptation (OTTA) offers a promising solution by enabling models to dynamically adjust to new data distributions without requiring source data or labeled target samples. Nevertheless, existing OTTA methods are largely designed for traditional artificial neural networks and are not well-suited for SNNs. To address this gap, we propose a low-power, neuromorphic chip-friendly online test-time adaptation framework, aiming to enhance model generalization under distribution shifts. The proposed approach is called Threshold Modulation (TM), which dynamically adjusts the firing threshold through neuronal dynamics-inspired normalization, being more compatible with neuromorphic hardware. Experimental results on benchmark datasets demonstrate the effectiveness of this method in improving the robustness of SNNs against distribution shifts while maintaining low computational cost. The proposed method offers a practical solution for online test-time adaptation of SNNs, providing inspiration for the design of future neuromorphic chips. The demo code is available at github.com/NneurotransmitterR/TM-OTTA-SNN.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes Threshold Modulation (TM), a framework for online test-time adaptation (OTTA) of spiking neural networks (SNNs) deployed on neuromorphic hardware. TM dynamically adjusts neuronal firing thresholds via a normalization step inspired by neuronal dynamics to mitigate distribution shifts without access to source data or target labels. The authors claim this yields improved robustness on benchmark datasets while preserving low computational cost and neuromorphic compatibility; demo code is provided.

Significance. If the normalization step can be realized with strictly local, event-driven spiking operations, the approach would address a genuine gap in hardware-friendly OTTA for SNNs and support more robust edge deployment. The explicit release of demo code is a clear strength that aids reproducibility and allows direct verification of the low-power claims.

major comments (2)
  1. [Method] Method section (normalization procedure): the description of 'neuronal dynamics-inspired normalization' does not specify how mean or variance statistics are computed or applied using only membrane potentials and spikes. This detail is load-bearing for the central hardware-compatibility claim, because any reliance on batch-wise averaging or auxiliary digital accumulators would violate the asynchronous, low-power constraints asserted in the abstract.
  2. [Experiments] Experimental results: no quantitative metrics, ablation tables, or baseline comparisons are supplied to support the statement that TM 'demonstrate[s] the effectiveness' under distribution shifts. Without these, the robustness claim cannot be evaluated and remains unverified.
minor comments (2)
  1. [Abstract] Abstract, final sentence of the TM description contains a dangling participial phrase ('being more compatible...') that should be rephrased for grammatical clarity.
  2. [Introduction] Ensure the introduction cites the most recent OTTA methods for ANNs so the claimed gap for SNNs is precisely delineated.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The comments have helped us identify areas where additional detail and experimental support are needed to strengthen the hardware-compatibility claims and empirical validation. We address each major comment below and have revised the manuscript accordingly.

read point-by-point responses

  1. Referee: [Method] Method section (normalization procedure): the description of 'neuronal dynamics-inspired normalization' does not specify how mean or variance statistics are computed or applied using only membrane potentials and spikes. This detail is load-bearing for the central hardware-compatibility claim, because any reliance on batch-wise averaging or auxiliary digital accumulators would violate the asynchronous, low-power constraints asserted in the abstract.

    Authors: We agree that the original Method section did not provide sufficient implementation detail on the local computation of statistics. In the revised manuscript we have expanded this section with an explicit description: mean and variance are maintained as per-neuron running estimates that are updated solely from each neuron's own membrane potential trajectory and recent spike count using an online, event-driven rule. No batch-wise operations or auxiliary digital accumulators are involved. We have added the corresponding equations, a local-update pseudocode listing, and a schematic diagram to make the strictly local, asynchronous nature unambiguous. revision: yes

  2. Referee: [Experiments] Experimental results: no quantitative metrics, ablation tables, or baseline comparisons are supplied to support the statement that TM 'demonstrate[s] the effectiveness' under distribution shifts. Without these, the robustness claim cannot be evaluated and remains unverified.

    Authors: The referee correctly notes that the initial submission contained only a high-level statement without supporting numbers or tables. We have now added a dedicated Experiments section that reports classification accuracy under multiple distribution-shift scenarios on standard benchmarks, includes ablation studies that isolate the contribution of the threshold-modulation component, and provides direct comparisons against both non-adaptive SNN baselines and adapted versions of existing OTTA methods. These additions supply the quantitative evidence needed to substantiate the robustness claims. revision: yes

Circularity Check

0 steps flagged

No circularity: method proposed as independent normalization technique

full rationale

The paper introduces Threshold Modulation (TM) as a new online test-time adaptation framework for SNNs that dynamically adjusts firing thresholds via neuronal dynamics-inspired normalization. This is presented as an original proposal compatible with neuromorphic hardware, supported by experimental results on benchmarks rather than any self-referential derivation. No equations, parameters, or claims reduce by construction to prior fits, self-citations, or renamed inputs. The derivation chain remains self-contained with independent content.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that threshold adjustment through dynamics-inspired normalization is both sufficient and hardware-compatible for adaptation; no explicit free parameters or new entities are named in the abstract.

axioms (1)
  • domain assumption Neuronal dynamics-inspired normalization can be applied to adjust firing thresholds in a way that counters distribution shifts
    Invoked as the core mechanism enabling adaptation without source data or labels.

pith-pipeline@v0.9.0 · 5764 in / 1106 out tokens · 49749 ms · 2026-05-22T15:55:09.723066+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages

  1. [1]

    Toward large-scale spiking neural networks: A comprehensive survey and future directions,

    Y . Hu, Q. Zheng, G. Li, H. Tang, and G. Pan, “Toward large-scale spiking neural networks: A comprehensive survey and future directions,” 2024,arXiv:2409.02111

    work page arXiv 2024

  2. [2]

    Effective transfer learning algorithm in spiking neural networks,

    Q. Zhan, G. Liu, X. Xie, G. Sun, and H. Tang, “Effective transfer learning algorithm in spiking neural networks,”IEEE Transactions on Cybernetics, vol. 52, no. 12, pp. 13 323–13 335, Dec. 2022

    work page 2022

  3. [3]

    Brain-inspired online adaptation for remote sensing with spiking neural network,

    D. Duan, P. Liu, B. Hui, and F. Wen, “Brain-inspired online adaptation for remote sensing with spiking neural network,”IEEE Transactions on Geoscience and Remote Sensing, Jan. 2025

    work page 2025

  4. [4]

    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,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 34, 2021, pp. 21 056– 21 069

    work page 2021

  5. [5]

    Con- version of continuous-valued deep networks to efficient event-driven networks for image classification,

    B. Rueckauer, I.-A. Lungu, Y . Hu, M. Pfeiffer, and S.-C. Liu, “Con- version of continuous-valued deep networks to efficient event-driven networks for image classification,”Frontiers in Neuroscience, vol. 11, p. 682, Dec. 2017

    work page 2017

  6. [6]

    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, May 2018

    work page 2018

  7. [7]

    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, Nov. 2019

    work page 2019

  8. [8]

    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. Jain, Y . Liao, C.-K. Lin, A. Lines, R. Liu, D. Mathaikutty, S. McCoy, A. Paul, J. Tse, G. Venkataramanan, Y .-H. Weng, A. Wild, Y . Yang, and H. Wang, “Loihi: A neuromorphic manycore processor with on-chip learning,”IEEE Micro, vol. 38, no. 1, pp. 82–99,...

    work page 2018

  9. [9]

    TrueNorth: Design and tool flow of a 65 mw 1 million neuron programmable neurosynaptic chip,

    F. Akopyan, J. Sawada, A. Cassidy, R. Alvarez-Icaza, J. Arthur, P. Merolla, N. Imam, Y . Nakamura, P. Datta, G.-J. Nam, B. Taba, M. Beakes, B. Brezzo, J. B. Kuang, R. Manohar, W. P. Risk, B. Jackson, and D. S. Modha, “TrueNorth: Design and tool flow of a 65 mw 1 million neuron programmable neurosynaptic chip,”IEEE Transactions on Computer-Aided Design of ...

    work page 2015

  10. [10]

    Spike-based dynamic computing with asynchronous sensing-computing neuromorphic chip,

    M. Yao, O. Richter, G. Zhao, N. Qiao, Y . Xing, D. Wang, T. Hu, W. Fang, T. Demirci, M. De Marchi, L. Deng, T. Yan, C. Nielsen, S. Sheik, C. Wu, Y . Tian, B. Xu, and G. Li, “Spike-based dynamic computing with asynchronous sensing-computing neuromorphic chip,” Nature Communications, vol. 15, p. 4464, May 2024

    work page 2024

  11. [11]

    Darwin3: A large-scale neuromorphic chip with a novel isa and on-chip learning,

    D. Ma, X. Jin, S. Sun, Y . Li, X. Wu, Y . Hu, F. Yang, H. Tang, X. Zhu, P. Lin, and G. Pan, “Darwin3: A large-scale neuromorphic chip with a novel isa and on-chip learning,”National Science Review, vol. 11, no. 5, p. nwae102, May 2024

    work page 2024

  12. [12]

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

    N. Qiao, H. Mostafa, F. Corradi, M. Osswald, F. Stefanini, D. Sum- islawska, and G. Indiveri, “A reconfigurable on-line learning spiking neuromorphic processor comprising 256 neurons and 128k synapses,” Frontiers in neuroscience, vol. 9, p. 141, Apr. 2015

    work page 2015

  13. [13]

    A 0.086-mm 2 12.7- pj/sop 64k-synapse 256-neuron online-learning digital spiking neuro- morphic processor in 28-nm cmos,

    C. Frenkel, M. Lefebvre, J.-D. Legat, and D. Bol, “A 0.086-mm 2 12.7- pj/sop 64k-synapse 256-neuron online-learning digital spiking neuro- morphic processor in 28-nm cmos,”IEEE Transactions on Biomedical Circuits and Systems, vol. 13, no. 1, pp. 145–158, Feb. 2019

    work page 2019

  14. [14]

    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, Jul. 2020

    work page 2020

  15. [15]

    ReckOn: A 28nm sub-mm2 task-agnostic spiking recurrent neural network processor enabling on-chip learning over second-long timescales,

    C. Frenkel and G. Indiveri, “ReckOn: A 28nm sub-mm2 task-agnostic spiking recurrent neural network processor enabling on-chip learning over second-long timescales,” inIEEE International Solid-State Circuits Conference (ISSCC), vol. 65, 2022, pp. 1–3

    work page 2022

  16. [16]

    E-prop on spinnaker 2: Exploring online learning in spiking rnns on neuromorphic hardware,

    A. Rostami, B. V ogginger, Y . Yan, and C. G. Mayr, “E-prop on spinnaker 2: Exploring online learning in spiking rnns on neuromorphic hardware,” Frontiers in Neuroscience, vol. 16, p. 1018006, Nov. 2022

    work page 2022

  17. [17]

    Tent: Fully test-time adaptation by entropy minimization,

    D. Wang, E. Shelhamer, S. Liu, B. Olshausen, and T. Darrell, “Tent: Fully test-time adaptation by entropy minimization,” inInternational Conference on Learning Representations (ICLR), 2021

    work page 2021

  18. [18]

    MECTA: Memory- Economic continual test-time model adaptation,

    J. Hong, L. Lyu, J. Zhou, and M. Spranger, “MECTA: Memory- Economic continual test-time model adaptation,” inInternational Con- ference on Learning Representations (ICLR), 2023

    work page 2023

  19. [19]

    Towards stable test-time adaptation in dynamic wild world,

    S. Niu, J. Wu, Y . Zhang, Z. Wen, Y . Chen, P. Zhao, and M. Tan, “Towards stable test-time adaptation in dynamic wild world,” inInternational Conference on Learning Representations (ICLR), 2023

    work page 2023

  20. [20]

    Parameter- free online test-time adaptation,

    M. Boudiaf, R. Mueller, I. B. Ayed, and L. Bertinetto, “Parameter- free online test-time adaptation,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 8334–8343

    work page 2022

  21. [21]

    Test-time adaptation via self- training with nearest neighbor information,

    M. Jang, S.-Y . Chung, and H. W. Chung, “Test-time adaptation via self- training with nearest neighbor information,” inInternational Conference on Learning Representations (ICLR), 2023

    work page 2023

  22. [22]

    Continual test-time domain adaptation,

    Q. Wang, O. Fink, L. Van Gool, and D. Dai, “Continual test-time domain adaptation,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 7191–7201

    work page 2022

  23. [23]

    Ef- ficient test-time model adaptation without forgetting,

    S. Niu, J. Wu, Y . Zhang, Y . Chen, S. Zheng, P. Zhao, and M. Tan, “Ef- ficient test-time model adaptation without forgetting,” inInternational Conference on Machine Learning (ICML), vol. 162, 2022, pp. 16 888– 16 905

    work page 2022

  24. [24]

    Batch Normalization: Accelerating deep network training by reducing internal covariate shift,

    S. Ioffe and C. Szegedy, “Batch Normalization: Accelerating deep network training by reducing internal covariate shift,” inInternational Conference on Machine Learning (ICML), vol. 37, 2015

    work page 2015

  25. [25]

    Evaluating prediction-time batch normalization for robustness under covariate shift

    Z. Nado, S. Padhy, D. Sculley, A. D’Amour, B. Lakshminarayanan, and J. Snoek, “Evaluating prediction-time batch normalization for robustness under covariate shift,” 2021,arXiv:2006.10963

    work page arXiv 2021

  26. [26]

    Improving robustness against common corruptions by co- variate shift adaptation,

    S. Schneider, E. Rusak, L. Eck, O. Bringmann, W. Brendel, and M. Bethge, “Improving robustness against common corruptions by co- variate shift adaptation,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 33, 2020, pp. 11 539–11 551

    work page 2020

  27. [27]

    Convolutional networks for fast, energy-efficient neuromorphic computing,

    S. K. Esser, P. A. Merolla, J. V . Arthur, A. S. Cassidy, R. Appuswamy, A. Andreopoulos, D. J. Berg, J. L. McKinstry, T. Melano, D. R. Barch, C. di Nolfo, P. Datta, A. Amir, B. Taba, M. D. Flickner, and D. S. Modha, “Convolutional networks for fast, energy-efficient neuromorphic computing,”Proceedings of the National Academy of Sciences, vol. 113, no. 41,...

    work page 2016

  28. [28]

    Membrane potential batch normalization for spiking neural networks,

    Y . Guo, Y . Zhang, Y . Chen, W. Peng, X. Liu, L. Zhang, X. Huang, and Z. Ma, “Membrane potential batch normalization for spiking neural networks,” inIEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 19 363–19 373

    work page 2023

  29. [29]

    LTMD: Learning improvement of spiking neural networks with learnable thresholding neurons and moderate dropout,

    S. Wang, T. H. Cheng, and M.-H. Lim, “LTMD: Learning improvement of spiking neural networks with learnable thresholding neurons and moderate dropout,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022, pp. 28 350–28 362

    work page 2022

  30. [30]

    DIET-SNN: A low-latency spiking neural network with direct input encoding and leakage and threshold optimization,

    N. Rathi and K. Roy, “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, vol. 34, no. 6, pp. 3174–3182, Jun. 2023

    work page 2023

  31. [31]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky and G. Hinton, “Learning multiple layers of features from tiny images,” 2009

    work page 2009

  32. [32]

    Benchmarking Neural Network Ro- bustness to Common Corruptions and Perturbations,

    D. Hendrycks and T. Dietterich, “Benchmarking Neural Network Ro- bustness to Common Corruptions and Perturbations,” inInternational Conference on Learning Representations (ICLR), 2019

    work page 2019

  33. [33]

    ImageNet large scale visual recognition challenge,

    O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein, A. C. Berg, and L. Fei- Fei, “ImageNet large scale visual recognition challenge,”International Journal of Computer Vision, vol. 115, pp. 211–252, Dec. 2015

    work page 2015

  34. [34]

    Neuro-modulated hebbian learning for fully test-time adaptation,

    Y . Tang, C. Zhang, H. Xu, S. Chen, J. Cheng, L. Leng, Q. Guo, and Z. He, “Neuro-modulated hebbian learning for fully test-time adaptation,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 3728–3738

    work page 2023

  35. [35]

    Reading digits in natural images with unsupervised feature learning,

    Y . Netzer, T. Wang, A. Coates, A. Bissacco, B. Wu, and A. Y . Ng, “Reading digits in natural images with unsupervised feature learning,” in NIPS Workshop on Deep Learning and Unsupervised Feature Learning, 2011

    work page 2011

  36. [36]

    Gradient-based learning applied to document recognition,

    Y . Lecun, L. Bottou, Y . Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,”Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, Nov. 1998

    work page 1998

  37. [37]

    Domain-adversarial training of neural networks,

    Y . Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Lavi- olette, M. March, and V . Lempitsky, “Domain-adversarial training of neural networks,”Journal of Machine Learning Research, vol. 17, no. 59, pp. 1–35, 2016

    work page 2016

  38. [38]

    A database for handwritten text recognition research,

    J. Hull, “A database for handwritten text recognition research,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 16, no. 5, pp. 550–554, May 1994

    work page 1994

  39. [39]

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

    Y . Wu, L. Deng, G. Li, J. Zhu, Y . Xie, and L. Shi, “Direct training for spiking neural networks: Faster, larger, better,” inAAAI Conference on Artificial Intelligence, vol. 33, no. 01, 2019, pp. 1311–1318

    work page 2019

  40. [40]

    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,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 34, 2021, pp. 23 426–23 439

    work page 2021

  41. [41]

    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,” inIEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 2641–2651

    work page 2021

  42. [42]

    Object detection with spiking neural networks on automotive event data,

    L. Cordone, B. Miramond, and P. Thierion, “Object detection with spiking neural networks on automotive event data,” inInternational Joint Conference on Neural Networks (IJCNN), 2022, pp. 1–8

    work page 2022

  43. [43]

    Convolutional spiking neural networks for spatio- temporal feature extraction,

    A. Samadzadeh, F. S. T. Far, A. Javadi, A. Nickabadi, and M. H. Chehreghani, “Convolutional spiking neural networks for spatio- temporal feature extraction,”Neural Processing Letters, vol. 55, pp. 6979–6995, Dec. 2023

    work page 2023

  44. [44]

    The norm must go on: Dynamic unsupervised domain adaptation by normalization,

    M. J. Mirza, J. Micorek, H. Possegger, and H. Bischof, “The norm must go on: Dynamic unsupervised domain adaptation by normalization,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 14 745–14 755

    work page 2022

  45. [45]

    NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation,

    W. Fang, Z. Yu, Y . Chen, T. Huang, T. Masquelier, and Y . Tian, “NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022, pp. 27 253–27 266

    work page 2022

  46. [46]

    Entropy is not enough for test-time adaptation: From the perspective of disentangled factors,

    J. Lee, D. Jung, S. Lee, J. Park, J. Shin, U. Hwang, and S. Yoon, “Entropy is not enough for test-time adaptation: From the perspective of disentangled factors,” inInternational Conference on Learning Rep- resentations (ICLR), 2024

    work page 2024

  47. [47]

    Fangwei123456/spikingjelly,

    fangwei123456, “Fangwei123456/spikingjelly,” Jan. 2025. [Online]. Available: https://github.com/fangwei123456/spikingjelly

    work page 2025

  48. [48]

    Aaron-xichen/pytorch-playground,

    A. Chen, “Aaron-xichen/pytorch-playground,” May 2020. [Online]. Available: https://github.com/aaron-xichen/pytorch-playground

    work page 2020

  49. [49]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” inInternational Conference on Learning Representations (ICLR), 2015

    work page 2015

  50. [50]

    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, Oct. 2023

    work page 2023

  51. [51]

    AugMix: A simple data processing method to improve robustness and uncertainty,

    D. Hendrycks, N. Mu, E. D. Cubuk, B. Zoph, J. Gilmer, and B. Lak- shminarayanan, “AugMix: A simple data processing method to improve robustness and uncertainty,” inInternational Conference on Learning Representations (ICLR), 2019

    work page 2019

  52. [52]

    1.1 computing’s energy problem (and what we can do about it),

    M. Horowitz, “1.1 computing’s energy problem (and what we can do about it),” inIEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC), 2014, pp. 10–14

    work page 2014

  53. [53]

    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,” inIEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 6143–6153

    work page 2023

  54. [54]

    Long short-term memory and Learning-to-learn in networks of spiking neu- rons,

    G. Bellec, D. Salaj, A. Subramoney, R. Legenstein, and W. Maass, “Long short-term memory and Learning-to-learn in networks of spiking neu- rons,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 31, 2018

    work page 2018