pith. sign in

arxiv: 2605.17686 · v1 · pith:U56PN4P4new · submitted 2026-05-17 · 💻 cs.CV

Brain-inspired spike-timing plasticity for reliable label-efficient event-camera vision

Mohamad Yazan Sadoun , Sarah Sharif , Yaser Mike Banad This is my paper

Pith reviewed 2026-05-20 13:07 UTC · model grok-4.3

classification 💻 cs.CV
keywords event cameraSTDPspike-timing-dependent plasticitylabel-efficientobject detectiondrift robustnessneuromorphic
0
0 comments X

The pith

Local STDP modules enable reliable label-efficient event-camera vision on a single CPU thread.

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

This paper introduces local spike-timing-dependent plasticity rules to address the high labeling and compute costs of event-camera object detectors. Three modules—sequence, candidate, and tube-reliability—run locally without GPUs or dense gradients. On the FRED benchmark these rules support zero-label detection at 53.8 percent mAP, scale to 76.9 percent with minimal supervision, and reach 78.6 percent with the candidate gate. The same gate improves robustness to acquisition drift over streaming k-means and lowers single-model variance substantially.

Core claim

The paper claims that its three local STDP modules allow a detector to operate across label budgets from zero to light supervision, achieving 53.8 percent mAP with no labels, 76.9 percent with roughly 26 bits of supervision, and 78.60 percent with the candidate-reliability gate on the FRED drone benchmark. Under acquisition-order drift the gate beats streaming k-means by 2.03 points in every one of 20 trials while a control without drift shows no such advantage. The approach also cuts performance variance by a factor of 6.6, matches a 44-seed ensemble with a single model, transfers to the Intel Lava simulator at 89 percent top-2 agreement, and lowers false alarms on the EVUAV benchmark.

What carries the argument

The candidate-reliability STDP gate, which applies local spike-timing rules to filter and stabilize detections without global training.

If this is right

  • The STDP candidate-reliability gate reaches 78.60 percent mAP@30 on FRED.
  • It outperforms streaming k-means by 2.03 points under acquisition-order drift in all 20 trials.
  • STDP reduces single-model variance by 6.6 times.
  • The gate transfers to Intel Lava with 89 percent top-2 agreement.
  • Tube-level STDP reduces false alarms on EVUAV at high detection probability.

Where Pith is reading between the lines

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

  • These local rules could enable real-time vision on low-power edge devices such as drones without extensive labeling.
  • The variance reduction may reduce the need for model ensembles in practical deployments.
  • The drift robustness suggests potential for continuous learning in changing environments.
  • Similar STDP modules might apply to other spiking sensor data beyond vision.

Load-bearing premise

That the performance gains and drift robustness are caused by the specific local STDP rules rather than by unstated implementation choices, dataset quirks, or post-hoc module tuning.

What would settle it

Replacing the STDP modules with non-timing-based selection rules in the same pipeline and checking whether the mAP improvement and drift outperformance disappear.

Figures

Figures reproduced from arXiv: 2605.17686 by Mohamad Yazan Sadoun, Sarah Sharif, Yaser Mike Banad.

Figure 1
Figure 1. Figure 1: Pipeline overview. a, ON/OFF event scatter for a drone in flight. b, Six classical event-detection channels feed an IoU-distinctness union; a 17-dimensional fingerprint drives a 384-plastic-synapse STDP-trained spiking cohort gate (17 → 32 → 12 LIF) that selects one of six processing recipes per cohort [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Streaming distribution-shift adaptation with a no-drift control. a, Streaming protocol: 184 FRED training sequences split by acquisition order into initial (n=113) and shifted (n=71) streams; the cohort gate is reassigned for a held-out 30-sequence evaluation slice after every update. b, Per-step Eval mAP@30, mean ± std across N=20 frozen￾Who seeds, five methods. c, Per-seed SNN advantage over streaming k-… view at source ↗
Figure 3
Figure 3. Figure 3: Log-log plot of mAP@30 standard deviation versus ensemble size N. The streaming-protocol analytic bound σrand/√N (with σrand=0.607 pp from N=19 architecture-only seeds; Tab. 3) intersects σSTDP=0.092 pp at N*=44. Empirical plurality-vote anchors approach this bound within 1.45× at N=44. STDP-Tube improves EV-UAV operating points EV-UAV [32] uses DAVIS346 sensors with few-pixel targets and reports event-lev… view at source ↗
Figure 4
Figure 4. Figure 4: Candidate-level reliability with STDP-Tube on EV-UAV. a, STDP-Tube links per-frame candidates into short event tubes and maps tube features to six output states (including an explicit unknown) via a bounded WTA/STDP reliability layer. b, EV-UAV operating points at locked defaults (Tab. 4): STDP-Tube K=5 (diamond) reaches Fa = 340.21×10−4 at Pd = 88.44%, below the six-channel baseline (Fa = 662.34) and the … view at source ↗
Figure 5
Figure 5. Figure 5: Six representative sequences across five drone types (DJI Mini 3, DJI Tello EDU, DJI Mini 2, BetaFPV Air75, DarwinFPV CineApe 20) plus a multi-drone scene. RGB ground-truth view and event-stream view shown side by side at the same instant; green dashed box = ground truth, orange solid = model prediction [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
read the original abstract

Deploying event-camera object detectors is constrained by per-frame labeling requirements and GPU compute demands. This work introduces three local spike-timing-dependent plasticity (STDP) modules, including sequence, candidate, and tube-reliability modules, that operate on a single CPU thread without GPU support. On the FRED drone benchmark, the proposed framework spans three label-efficient supervision tiers. A strict zero-label detector achieves 53.8% mAP@30, approximately 26 train-derived bits achieve 76.9% mAP@30, and an STDP candidate-reliability gate achieves 78.60 +/- 0.42% mAP@30. Under acquisition-order drift, the cohort gate outperforms streaming k-means by 2.03 +/- 0.58 percentage points across 20 of 20 positive trials, while a no-drift control falsifies the effect. STDP reduces single-model variance by 6.6 times, and one trained gate matches a 44-seed ensemble bound. The gate transfers to Intel Lava with 89% top-2 agreement. On the EVUAV benchmark, a tube-level STDP layer reduces false alarms from 454 to 331e-4 at Pd >= 88%. Dense gradient-trained detectors cannot provide this combination of gradient training, dense matrix multiplication, and local plasticity-free operation by construction.

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

1 major / 3 minor

Summary. The paper introduces three local spike-timing-dependent plasticity (STDP) modules (sequence, candidate, and tube-reliability) for label-efficient object detection with event cameras, running on a single CPU thread without GPU. On the FRED benchmark it reports a zero-label detector at 53.8% mAP@30, ~26 train-derived bits reaching 76.9% mAP@30, and an STDP candidate-reliability gate at 78.60 ± 0.42% mAP@30; under acquisition-order drift the gate outperforms streaming k-means by 2.03 ± 0.58 points in all 20 trials, reduces single-model variance by 6.6×, and transfers to Intel Lava with 89% top-2 agreement. A tube-level STDP layer on EVUAV reduces false alarms from 454 to 331×10^{-4} at Pd ≥ 88%.

Significance. If the attribution to the local STDP rules holds, the work demonstrates a practical route to label-efficient, drift-robust event-camera detection on CPU-only hardware, which would be valuable for resource-constrained applications. The reported error bars, 20/20 trial consistency, and explicit no-drift control provide concrete support for the drift-robustness claim; the Lava transfer result further indicates hardware portability.

major comments (1)
  1. [Results and Methods] Results and Methods sections: the central attribution of the 78.60% mAP@30, 2.03-point drift gain, and 6.6× variance reduction to the three local STDP modules is not isolated by any ablation that disables only the STDP update rules while preserving the remainder of the single-CPU pipeline, feature extraction, and candidate-selection logic. Without such controls it remains possible that the gains arise from unstated implementation choices rather than the plasticity equations themselves.
minor comments (3)
  1. [Abstract] Abstract and Methods: the phrase 'approximately 26 train-derived bits' is not defined; a precise description of how these bits are obtained and counted would clarify the supervision tiers.
  2. [Methods] Methods: data splits, exact values of the STDP timing windows and reliability thresholds, and the procedure for selecting the 20 acquisition-order trials are not reported, limiting reproducibility.
  3. [Figures] Figure captions: axis labels and error-bar definitions should be stated explicitly rather than relying on the main text.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The major comment raises a valid point about the need for more precise isolation of the STDP modules' contributions. We agree that this would strengthen the attribution and will incorporate the requested controls in the revised manuscript.

read point-by-point responses

  1. Referee: [Results and Methods] Results and Methods sections: the central attribution of the 78.60% mAP@30, 2.03-point drift gain, and 6.6× variance reduction to the three local STDP modules is not isolated by any ablation that disables only the STDP update rules while preserving the remainder of the single-CPU pipeline, feature extraction, and candidate-selection logic. Without such controls it remains possible that the gains arise from unstated implementation choices rather than the plasticity equations themselves.

    Authors: We acknowledge that the manuscript does not include an ablation that disables only the STDP update rules while keeping feature extraction, candidate-selection logic, and the single-CPU pipeline unchanged. The existing comparisons to streaming k-means under drift conditions (including the no-drift control and 20/20 trial consistency) serve as a non-plasticity baseline for the overall system, but they do not isolate the plasticity equations themselves. To address this directly, we will add a targeted ablation study in the revised Results and Methods sections. In this ablation, the STDP update rules will be disabled (e.g., by freezing weights after initialization or replacing with non-plastic equivalents) while preserving all other components. We will report the resulting mAP@30, drift gain, and variance metrics to clarify the specific contribution of the local STDP modules. revision: yes

Circularity Check

0 steps flagged

No circularity detected; claims rest on empirical benchmarks

full rationale

The manuscript reports measured mAP@30 values, drift-robustness deltas versus streaming k-means, variance reductions, and cross-platform transfer on FRED and EVUAV benchmarks. These are presented as experimental outcomes from CPU-thread implementations of the three STDP modules, with explicit controls (no-drift, 20-trial consistency) and external baselines. No equations, self-definitions, or self-citations are shown that reduce any reported performance figure to a fitted parameter or prior result by construction. The final sentence notes a contrast with gradient-trained detectors 'by construction,' but this is a negative claim about alternatives rather than a self-referential derivation of the positive results.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Approach rests on standard STDP timing rules plus custom module definitions; a modest number of timing windows or reliability thresholds are likely chosen or fitted to achieve the quoted mAP values.

free parameters (1)
  • STDP timing windows and reliability thresholds
    These control when spikes are considered coincident or reliable and are expected to be set to match the reported performance on FRED and EVUAV.
axioms (1)
  • domain assumption Local spike-timing rules suffice to produce reliable object detection without global gradient updates
    Invoked by the design of the three modules and the claim that dense gradient training is unnecessary.

pith-pipeline@v0.9.0 · 5776 in / 1385 out tokens · 99603 ms · 2026-05-20T13:07:29.791088+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

49 extracted references · 49 canonical work pages · 1 internal anchor

  1. [1]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 44(1), 154–180 (2022)

    Gallego, G., Delbrück, T., Orchard, G., Bartolozzi, C., Taba, B., Censi, A., Leutenegger, S., Davison, A.J., Conradt, J., Daniilidis, K., Scaramuzza, D.: Event-based vision: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence 44(1), 154–180 (2022)

    work page 2022

  2. [2]

    Deep learning for event-based vision: A comprehensive survey and benchmarks.arXiv preprint arXiv:2302.08890,

    Zheng, X., Liu, Y., Lu, Y., Hua, T., Pan, T., Zhang, W., Tao, D., Wang, L.: Deep learning for event-based vision: A comprehensive survey and benchmarks. arXiv:2302.08890 (2024)

    work page arXiv 2024

  3. [3]

    Journal of Neuroscience 18(24), 10464–10472 (1998)

    Bi, G.-q., Poo, M.-m.: Synaptic modifications in cultured hippocampal neurons: Dependence on spike timing, synaptic strength, and postsynaptic cell type. Journal of Neuroscience 18(24), 10464–10472 (1998)

    work page 1998

  4. [4]

    Cambridge University Press, Cambridge, UK (2002)

    Gerstner, W., Kistler, W.M.: Spiking Neuron Models: Single Neurons, Populations, Plasticity. Cambridge University Press, Cambridge, UK (2002)

    work page 2002

  5. [5]

    Neuromorphic Computing and Engineering (2024)

    Ahmadvand, R., Sharif, S.S., Banad, Y.M.: A cloud-edge framework for energy-efficient event-driven control: an integration of online supervised learning, spiking neural networks and local plasticity rules. Neuromorphic Computing and Engineering (2024)

    work page 2024

  6. [6]

    Scientific Reports (2025)

    Ahmadvand, R., Sharif, S.S., Banad, Y.M.: Neuromorphic robust framework for integrated estimation and control in dynamical systems using spiking neural networks. Scientific Reports (2025)

    work page 2025

  7. [7]

    arXiv:2507.00443 (2025)

    Ahmadvand, R., Sharif, S.S., Banad, Y.M.: Novel pigeon-inspired 3D obstacle detection and avoidance maneuver for multi- UAV systems. arXiv:2507.00443 (2025)

    work page arXiv 2025

  8. [8]

    IEEE Micro 38(1), 82–99 (2018)

    Davies, M., Srinivasa, N., Lin, T.-H., et al.: Loihi: A neuromorphic manycore processor with on-chip learning. IEEE Micro 38(1), 82–99 (2018)

    work page 2018

  9. [9]

    Proceedings of the IEEE 109(5), 911–934 (2021)

    Davies, M., Wild, A., Orchard, G., et al.: Advancing neuromorphic computing with Loihi: A survey of results and outlook. Proceedings of the IEEE 109(5), 911–934 (2021)

    work page 2021

  10. [10]

    Intel technology brief

    Intel Labs: Taking Neuromorphic Computing with Loihi 2 to the Next Level. Intel technology brief. Accessed 2026-05-14 (2021)

    work page 2026

  11. [11]

    arXiv:2111.03746 (2021)

    Orchard, G., Frady, E.P., Rubin, D.B.D., Sanborn, S., Shrestha, S.B., Sommer, F.T., Davies, M.: Efficient neuromorphic signal processing with Loihi 2. arXiv:2111.03746 (2021)

    work page arXiv 2021

  12. [12]

    Proceedings of the IEEE 102(5), 652–665 (2014)

    Furber, S.B., Galluppi, F., Temple, S., Plana, L.A.: The SpiNNaker project. Proceedings of the IEEE 102(5), 652–665 (2014)

    work page 2014

  13. [13]

    IEEE Open Journal of Circuits and Systems (2025)

    Motaman, M., Sharif, S.S., Banad, Y.M.: Biologically-inspired, ultra-low power, and high-speed integrate-and-fire neuron circuit with stochastic behavior using nanoscale side-contacted field effect diode technology. IEEE Open Journal of Circuits and Systems (2025)

    work page 2025

  14. [14]

    IEEE Int

    Sadoun, M.Y., Sharif, S.S., Banad, Y.M.: How can neuromorphic hardware achieve energy-efficient CNN inference for edge AI? In: Proc. IEEE Int. Conf. on Artificial Intelligence × Data Knowledge Engineering (AIxDKE) (2026)

    work page 2026

  15. [15]

    Advanced Intelligent Discovery (2026)

    Larsh, L., Siddique, R., Sharif, S.S., Banad, Y.M.: Parametric analysis of spiking neurons in 16 nm fin field-effect transistor technology. Advanced Intelligent Discovery (2026)

    work page 2026

  16. [16]

    In: Proc

    Magrini, G., Marini, N., Becattini, F., Berlincioni, L., Biondi, N., Pala, P., Del Bimbo, A.: FRED: The Florence RGB-event drone dataset. In: Proc. ACM International Conference on Multimedia (ACM MM) (2025)

    work page 2025

  17. [17]

    In: Proc

    Gehrig, M., Scaramuzza, D.: Recurrent vision transformers for object detection with event cameras. In: Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR) (2023)

    work page 2023

  18. [18]

    In: Proc

    Hamaguchi, R., Furukawa, Y., Onishi, M., Sakurada, K.: Hierarchical neural memory network for low latency event processing. In: Proc. IEEE/CVF CVPR (2023)

    work page 2023

  19. [19]

    Nature 629, 1034–1040 (2024)

    Gehrig, D., Scaramuzza, D.: Low-latency automotive vision with event cameras. Nature 629, 1034–1040 (2024)

    work page 2024

  20. [20]

    Frontiers in Neuroscience 18 (2025)

    Silva, D.A., Smagulova, K., Elsheikh, A., Fouda, M.E., Eltawil, A.M.: A recurrent YOLOv8-based framework for event-based object detection. Frontiers in Neuroscience 18 (2025)

    work page 2025

  21. [21]

    In: Proc

    Niu, D., Yang, W., Yang, W., Bi, D., Ma, S., Wu, J.: AHM-Net: An asymmetric hierarchical multi-modal fusion network for robust UAV detection using RGB and event data. In: Proc. IEEE ICASSP (2026)

    work page 2026

  22. [22]

    arXiv:2603.08386

    Bezick, M., Sahin, M.: Real-Time Drone Detection in Event Cameras via Per-Pixel Frequency Analysis (2026). arXiv:2603.08386

    work page arXiv 2026

  23. [23]

    In: Proc

    Kim, S., Park, S., Na, B., Yoon, S.: Spiking-YOLO: Spiking neural network for energy-efficient object detection. In: Proc. AAAI (2020)

    work page 2020

  24. [24]

    In: Proc

    Su, Q., Chou, Y., Hu, Y., Li, J., Mei, S., Zhang, Z., Li, G.: Deep directly-trained spiking neural networks for object detection. In: Proc. IEEE/CVF ICCV (2023)

    work page 2023

  25. [25]

    In: Proc

    Luo, X., Yao, M., Chou, Y., Xu, B., Li, G.: Integer-valued training and spike-driven inference spiking neural network for high-performance and energy-efficient object detection. In: Proc. ECCV (2024)

    work page 2024

  26. [26]

    In: Proc

    Wang, Z., Wang, Z., Li, H., Qin, L., Jiang, R., Ma, D., Tang, H.: EAS-SNN: End-to-end adaptive sampling and representation for event-based detection with recurrent spiking neural networks. In: Proc. ECCV (2024)

    work page 2024

  27. [27]

    Neural Computation 3(1), 79–87 (1991)

    Jacobs, R.A., Jordan, M.I., Nowlan, S.J., Hinton, G.E.: Adaptive mixtures of local experts. Neural Computation 3(1), 79–87 (1991)

    work page 1991

  28. [28]

    JMLR 23(120), 1–39 (2022)

    Fedus, W., Zoph, B., Shazeer, N.: Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. JMLR 23(120), 1–39 (2022)

    work page 2022

  29. [29]

    Frontiers in Computational Neuroscience 9 (2015)

    Diehl, P.U., Cook, M.: Unsupervised learning of digit recognition using spike-timing-dependent plasticity. Frontiers in Computational Neuroscience 9 (2015)

    work page 2015

  30. [30]

    IEEE Transactions on Nanotechnology 12(3), 288–295 (2013)

    Querlioz, D., Bichler, O., Dollfus, P., Gamrat, C.: Immunity to device variations in a spiking neural network with memristive nanodevices. IEEE Transactions on Nanotechnology 12(3), 288–295 (2013)

    work page 2013

  31. [31]

    PLoS Computational Biology 9(4), 1003037 (2013)

    Nessler, B., Pfeiffer, M., Buesing, L., Maass, W.: Bayesian computation emerges in generic cortical microcircuits through spike-timing-dependent plasticity. PLoS Computational Biology 9(4), 1003037 (2013)

    work page 2013

  32. [32]

    In: Proc

    Chen, N., Xiao, C., Dai, Y., He, S., Li, M., An, W.: Event-based tiny object detection: A benchmark dataset and baseline. In: Proc. IEEE/CVF ICCV (2025)

    work page 2025

  33. [33]

    Benchmarking Keyword Spotting Efficiency on Neuromorphic Hardware

    Blouw, P., Choo, X., Hunsberger, E., Eliasmith, C.: Benchmarking keyword spotting efficiency on neuromorphic hardware. In: Proc. NICE (2019). arXiv:1812.01739

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  34. [34]

    Cell 135(3), 422–435 (2008)

    Turrigiano, G.G.: The self-tuning neuron: Synaptic scaling of excitatory synapses. Cell 135(3), 422–435 (2008)

    work page 2008

  35. [35]

    Triesch, J.: A gradient rule for the plasticity of a neuron's intrinsic excitability. Proc. ICANN, 65–70 (2005)

    work page 2005

  36. [36]

    Neural Computation 12(11), 2519–2535 (2000)

    Maass, W.: On the computational power of winner-take-all. Neural Computation 12(11), 2519–2535 (2000)

    work page 2000

  37. [37]

    In: Proc

    Boudiaf, M., Mueller, R., Ben Ayed, I., Bertinetto, L.: Parameter-free online test-time adaptation. In: Proc. IEEE/CVF CVPR (2022)

    work page 2022

  38. [38]

    In: Proc

    Liang, J., Hu, D., Feng, J.: Do we really need to access the source data? Source hypothesis transfer for unsupervised domain adaptation. In: Proc. ICML (2020)

    work page 2020

  39. [39]

    Pattern Recognition 80, 109–117 (2018)

    Li, Y., Wang, N., Shi, J., Liu, J., Hou, X.: Adaptive batch normalization for practical domain adaptation. Pattern Recognition 80, 109–117 (2018)

    work page 2018

  40. [40]

    The Annals of Statistics 7(1), 1–26 (1979)

    Efron, B.: Bootstrap methods: Another look at the jackknife. The Annals of Statistics 7(1), 1–26 (1979)

    work page 1979

  41. [41]

    In: Proc

    Ester, M., Kriegel, H.-P., Sander, J., Xu, X.: A density-based algorithm for discovering clusters in large spatial databases with noise. In: Proc. KDD, 226–231 (1996)

    work page 1996

  42. [42]

    IEEE Transactions on Information Theory 28(2), 129–137 (1982)

    Lloyd, S.P.: Least squares quantization in PCM. IEEE Transactions on Information Theory 28(2), 129–137 (1982)

    work page 1982

  43. [43]

    In: Proc

    Sculley, D.: Web-scale k-means clustering. In: Proc. WWW, 1177–1178 (2010)

    work page 2010

  44. [44]

    JMLR 12, 2825–2830 (2011)

    Pedregosa, F., et al.: Scikit-learn: Machine learning in Python. JMLR 12, 2825–2830 (2011)

    work page 2011

  45. [45]

    Journal of Pharmacokinetics and Biopharmaceutics 15(6), 657–680 (1987)

    Schuirmann, D.J.: A comparison of the two one-sided tests procedure and the power approach for assessing the equivalence of average bioavailability. Journal of Pharmacokinetics and Biopharmaceutics 15(6), 657–680 (1987)

    work page 1987

  46. [46]

    Nature Methods 17(3), 261– 272 (2020)

    Virtanen, P., et al.: SciPy 1.0: Fundamental algorithms for scientific computing in Python. Nature Methods 17(3), 261– 272 (2020)

    work page 2020

  47. [47]

    Scandinavian Journal of Statistics 6(2), 65–70 (1979)

    Holm, S.: A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics 6(2), 65–70 (1979)

    work page 1979

  48. [48]

    Version 0.10.0

    Intel Labs: Lava Software Framework. Version 0.10.0. https://github.com/lava-nc/lava (2021–2026)

    work page 2021

  49. [49]

    In: NeurIPS, 8024–8035 (2019)

    Paszke, A., et al.: PyTorch: An imperative style, high-performance deep learning library. In: NeurIPS, 8024–8035 (2019). Supplementary Information S1. Complete hyperparameter listing Table S1 enumerates every model and protocol hyperparameter referenced in Methods: spiking cohort gate architecture, STDP plasticity rule, six-channel detection pipeline, coh...

    work page 2019