pith. sign in

arxiv: 2509.14447 · v3 · submitted 2025-09-17 · 📡 eess.SP

Dual-Timescale Hebbian Accumulators for Online Spiking Neural Network Decoding in Intracortical Brain Machine Interfaces

Sriram V.C. Nallani , Sahil Shah This is my paper

Pith reviewed 2026-05-18 15:41 UTC · model grok-4.3

classification 📡 eess.SP
keywords spiking neural networksHebbian learningbrain-machine interfacesonline decodingeligibility tracesneural signal processinglow-memory trainingsupervised updates
0
0 comments X

The pith

A dual-timescale Hebbian accumulator enables online learning in spiking neural networks for brain-machine interfaces with memory use that does not grow with sequence length.

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

The paper proposes a new learning rule called the dual-timescale Hebbian accumulator for training spiking neural networks to decode signals from the brain in real time. This rule uses two different time scales of eligibility traces at each synapse, combined with error signals, to make updates at every time step while keeping the memory needed independent of the length of the data sequence. In tests on data from primate brain recordings, it reaches Pearson correlations of 0.81 or better on one set and 0.63 on another, while cutting memory use by 63 to 86 percent compared to standard backpropagation through time. Closed-loop tests also show it can adapt to disruptions in the neural signals and start learning without prior offline training.

Core claim

The author establishes that a dual-timescale Hebbian accumulator learning rule, which integrates synapse-specific fast and slow eligibility traces with error-modulated three-factor updates and integer-friendly RMS homeostasis, permits per-timestep online supervised learning in spiking neural networks for intracortical brain-machine interface decoding. This approach eliminates the requirement for backpropagation through time and maintains constant training memory regardless of sequence length. On two primate datasets, it delivers Pearson correlations of R greater than or equal to 0.81 on MC Maze and R greater than or equal to 0.63 on Zenodo Indy, accompanied by substantial memory savings of

What carries the argument

The dual-timescale Hebbian accumulator, consisting of fast and slow synapse-specific eligibility traces updated via three-factor Hebbian rules with error modulation and RMS normalization for homeostasis.

Load-bearing premise

The combination of fast and slow eligibility traces with error modulation and RMS homeostasis will stay stable and perform well even when neural signals change over time in real use, without extra optimizers or buffers.

What would settle it

Observing that the Pearson correlation drops below 0.5 or that memory usage begins to scale with sequence length during extended closed-loop operation with natural neural variations would indicate the claim does not hold.

Figures

Figures reproduced from arXiv: 2509.14447 by Sahil Shah, Sriram V.C. Nallani.

Figure 1
Figure 1. Figure 1: Proposed SNN architecture with recur￾rent connections and 2D velocity output. Our network is built around Leaky Integrate-and-Fire (LIF) neurons, chosen for their balance of biological realism and computational tractability (see Appendix D for background). Each neuron’s membrane potential ui(t) integrates incoming synaptic currents, leaks with a decay factor β, and emits a spike when reaching a threshold (… view at source ↗
Figure 2
Figure 2. Figure 2: Decoder performance: Pearson R, mean ± SEM, n = 10. Order of Decoders in (a) and (b): True Online SNN, Batched Online SNN, Kalman Filter, LSTM, BPTT SNN. MC Maze decoder comparison shown in (a), Zenodo Indy decoder comparison shown in (b). Performance of True (Timestep-wise) Online SNN on Zenodo Indy session 20161013_03 shown in (c). In figures (a) and (b), red significance stars indicate comparison with t… view at source ↗
Figure 3
Figure 3. Figure 3: Zenodo Indy ablation study: update rules, recurrence, RMS, timescales, mechanisms, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Learning curves: validation R vs. samples. Mean [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Closed-loop adaptation to disruptions (a-c) and learning from scratch (d). Mean over 10 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Complete per-timestep update procedure showing O(1) memory scaling. [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: MC Maze ablation study: update rules, recurrence, RMS, timescales, mechanisms, [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Memory efficiency comparison between Online SNN and BPTT SNN approaches for two [PITH_FULL_IMAGE:figures/full_fig_p020_8.png] view at source ↗
read the original abstract

Intracortical brain-machine interfaces require decoders that adapt continuously to neural signal instability while operating within strict memory budgets. We introduce a dual-timescale Hebbian accumulator learning rule for spiking neural networks that enables per-timestep online supervised updates with training memory constant in sequence length, avoiding backpropagation through time. The rule combines synapse-specific fast and slow eligibility traces, error-modulated three-factor updates, and integer-friendly RMS homeostasis, operating without adaptive gradient optimizers (Adam, RMSProp) or replay buffers. On two primate intracortical datasets, the method achieves Pearson correlations of $R \geq 0.81$ on MC~Maze and $R \geq 0.63$ on Zenodo~Indy, with 63--86\% measured memory reduction versus BPTT at sequence length $T = 1000$. Closed-loop simulations demonstrate online adaptation to neural disruptions and learning from scratch without offline calibration.

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 manuscript introduces a dual-timescale Hebbian accumulator learning rule for spiking neural networks in intracortical brain-machine interfaces. The rule enables per-timestep online supervised updates with memory constant in sequence length by combining synapse-specific fast and slow eligibility traces, error-modulated three-factor updates, and integer-friendly RMS homeostasis, avoiding backpropagation through time and replay buffers. On MC Maze and Zenodo Indy primate datasets it reports Pearson correlations R ≥ 0.81 and R ≥ 0.63 respectively, 63–86% memory reduction versus BPTT at T=1000, and successful closed-loop adaptation to neural disruptions without offline calibration.

Significance. If the performance and stability claims hold under rigorous verification, the approach could meaningfully advance practical BMI decoder design by addressing neural drift with low-memory online learning. The explicit memory scaling (O(1) in T) and avoidance of BPTT or adaptive optimizers are attractive for implantable hardware constraints, and the closed-loop results on real primate data provide direct empirical grounding for the online-adaptation claim.

major comments (2)
  1. [Results and closed-loop simulations] Results (closed-loop and offline evaluations): the reported correlations and memory reductions are given as point values without error bars, standard deviations across sessions or folds, or statistical tests against BPTT and other baselines; this weakens the ability to assess whether the 63–86% memory saving and R thresholds are reliably achieved or sensitive to post-hoc hyperparameter choices.
  2. [§3 (rule derivation)] Method (§3, eligibility trace and homeostasis equations): the stability claim for the combined fast/slow traces plus RMS homeostasis under real neural drift rests on the specific time-constant choices and the integer-friendly RMS term; the manuscript should include a sensitivity analysis or ablation removing the slow trace or the homeostasis term to confirm these components are load-bearing for the reported closed-loop performance.
minor comments (2)
  1. [Notation and §3] Clarify in the notation section whether the fast and slow eligibility traces are updated with exact integer arithmetic or require fixed-point approximations, and provide the precise update equations for the three-factor modulation.
  2. [Experimental results] In the memory-reduction figure or table, explicitly state the measured peak memory (in bytes or parameters) for both the proposed method and BPTT at T=1000 so the 63–86% figure can be reproduced.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback and recommendation for minor revision. We address each major comment point by point below, providing the strongest honest defense of the manuscript while incorporating revisions where the comments identify opportunities to strengthen the presentation.

read point-by-point responses

  1. Referee: [Results and closed-loop simulations] Results (closed-loop and offline evaluations): the reported correlations and memory reductions are given as point values without error bars, standard deviations across sessions or folds, or statistical tests against BPTT and other baselines; this weakens the ability to assess whether the 63–86% memory saving and R thresholds are reliably achieved or sensitive to post-hoc hyperparameter choices.

    Authors: We agree that variability measures and statistical comparisons improve the ability to evaluate reliability. In the revised manuscript we now report standard deviations across the multiple sessions available in each primate dataset, include error bars on the correlation and memory-reduction figures, and add Wilcoxon signed-rank tests against the BPTT baseline with associated p-values. These additions appear in the updated Results section and supplementary tables. We note that the original point estimates were obtained under the exact hyperparameter settings described in the methods; the new statistical tests confirm that the reported thresholds remain statistically distinguishable from the baseline even after accounting for session-to-session variability. revision: yes

  2. Referee: [§3 (rule derivation)] Method (§3, eligibility trace and homeostasis equations): the stability claim for the combined fast/slow traces plus RMS homeostasis under real neural drift rests on the specific time-constant choices and the integer-friendly RMS term; the manuscript should include a sensitivity analysis or ablation removing the slow trace or the homeostasis term to confirm these components are load-bearing for the reported closed-loop performance.

    Authors: We accept that explicit ablations strengthen the mechanistic claims. We have added a new subsection (4.3) containing two ablation experiments: (i) removal of the slow eligibility trace while retaining the fast trace and RMS homeostasis, and (ii) removal of the RMS homeostasis term while retaining both traces. In both cases closed-loop correlation drops substantially under the same neural-drift protocol, confirming that each component contributes to the observed stability. We have also included a limited sensitivity sweep of the fast and slow time constants (±20 % around the reported values) showing that performance remains within 5 % of the nominal R values. These results are now summarized in the main text and detailed in the supplement. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces a dual-timescale Hebbian accumulator learning rule as an original construction that combines synapse-specific fast and slow eligibility traces, error-modulated three-factor updates, and integer-friendly RMS homeostasis. This rule is defined directly in the manuscript and yields per-timestep online updates with memory independent of sequence length T by design, without reducing to fitted parameters from the evaluation data or relying on load-bearing self-citations for uniqueness theorems. Reported Pearson correlations and memory reductions are empirical measurements on external primate datasets (MC Maze, Zenodo Indy) rather than quantities forced by construction from the same inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that the proposed rule can be implemented in spiking networks and will generalize from the two primate datasets to ongoing neural instability; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption Spiking neural networks with the described eligibility traces and homeostasis can decode motor intent from intracortical recordings under non-stationary conditions.
    Invoked to justify the decoder's applicability to real BMI data.

pith-pipeline@v0.9.0 · 5701 in / 1331 out tokens · 47197 ms · 2026-05-18T15:41:05.811227+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

65 extracted references · 65 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

    work page

  2. [2]

    L. F. Abbott and S. B. Nelson. Synaptic plasticity: Taming the beast. Nature Neuroscience, 3 0 (11S): 0 1178--1183, 2000. doi:10.1038/81453

    work page doi:10.1038/81453 2000

  3. [3]

    Learning to learn by gradient descent by gradient descent

    Marcin Andrychowicz, Misha Denil, Sergio Gomez, Matthew Hoffman, David Pfau, Tom Schaul, Brendan Shillingford, and Nando de Freitas. Learning to learn by gradient descent by gradient descent. In Advances in Neural Information Processing Systems 29 (NeurIPS 2016), pp.\ 3981--3989, 2016

    work page 2016

  4. [4]

    Spiking neural network integrated circuits: A review of trends and future directions

    Arindam Basu, Charlotte Frenkel, Lei Deng, and Xueyong Zhang. Spiking neural network integrated circuits: A review of trends and future directions. arXiv preprint, 2022. Authors listed alphabetically

    work page 2022

  5. [5]

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

    Guillaume Bellec, Franz Scherr, Darjan Salaj, et al. A solution to the learning dilemma for recurrent networks of spiking neurons. Nature Communications, 11: 0 3625, 2020. doi:10.1038/s41467-020-17236-y

    work page doi:10.1038/s41467-020-17236-y 2020

  6. [6]

    Benna and Stefano Fusi

    Marcus K. Benna and Stefano Fusi. Computational principles of synaptic memory consolidation. Nature Neuroscience, 19 0 (12): 0 1697--1706, 2016

    work page 2016

  7. [7]

    Spiking neural classifier with lumped dendritic nonlinearity and binary synapses: A current mode vlsi implementation and analysis

    Abhishek Bhaduri, Amitava Banerjee, Subhrajit Roy, Shantanu Kar, and Arindam Basu. Spiking neural classifier with lumped dendritic nonlinearity and binary synapses: A current mode vlsi implementation and analysis. Neural Computation, 30 0 (3): 0 723--760, 2018. doi:10.1162/neco_a_01045

    work page doi:10.1162/neco_a_01045 2018

  8. [8]

    Synaptic modifications in cultured hippocampal neurons: Dependence on spike timing, synaptic strength, and postsynaptic cell type

    Guo-qiang Bi and Mu-ming Poo. Synaptic modifications in cultured hippocampal neurons: Dependence on spike timing, synaptic strength, and postsynaptic cell type. Journal of Neuroscience, 18 0 (24): 0 10464--10472, 1998

    work page 1998

  9. [9]

    Spike timing-dependent plasticity: A hebbian learning rule

    Natalia Caporale and Yang Dan. Spike timing-dependent plasticity: A hebbian learning rule. Annual Review of Neuroscience, 31: 0 25--46, 2008

    work page 2008

  10. [10]

    Jose M. Carmena. Advances in neuroprosthetic learning and control. PLoS Biology, 11 0 (5): 0 e1001561, 2013

    work page 2013

  11. [11]

    Mc\_maze: macaque primary motor and dorsal premotor cortex spiking activity during delayed reaching (version 0.220113.0400) [data set], 2022

    Mark Churchland and Matthew Kaufman. Mc\_maze: macaque primary motor and dorsal premotor cortex spiking activity during delayed reaching (version 0.220113.0400) [data set], 2022. URL https://doi.org/10.48324/dandi.000128/0.220113.0400

    work page doi:10.48324/dandi.000128/0.220113.0400 2022

  12. [12]

    Cunningham, Paul Nuyujukian, Vikash Gilja, Cindy A

    John P. Cunningham, Paul Nuyujukian, Vikash Gilja, Cindy A. Chestek, Stephen I. Ryu, and Krishna V. Shenoy. A closed-loop human simulator for investigating the role of feedback control in brain--machine interfaces. Journal of Neurophysiology, 105 0 (4): 0 1932--1949, 2011. doi:10.1152/jn.00503.2010

    work page doi:10.1152/jn.00503.2010 1932

  13. [13]

    Moorman, Amy L

    Siddharth Dangi, Suma Gowda, Hamilton G. Moorman, Amy L. Orsborn, Kelvin So, Maryam Shanechi, and Jose M. Carmena. Continuous closed-loop decoder adaptation with a recursive maximum likelihood algorithm allows for rapid performance acquisition in brain--machine interfaces. Neural Computation, 26 0 (9): 0 1811--1839, 2014

    work page 2014

  14. [14]

    Loihi: A neuromorphic manycore processor with on-chip learning

    Mike Davies, Narayan Srinivasa, Tien-Hsin Lin, et al. Loihi: A neuromorphic manycore processor with on-chip learning. IEEE Micro, 38 0 (1): 0 82--99, 2018

    work page 2018

  15. [15]

    Model-agnostic meta-learning for fast adaptation of deep networks

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model-agnostic meta-learning for fast adaptation of deep networks. In Proceedings of the 34th International Conference on Machine Learning (ICML 2017), volume 70 of Proceedings of Machine Learning Research, pp.\ 1126--1135, 2017

    work page 2017

  16. [16]

    R zvan V. Florian. Reinforcement learning through modulation of spike-timing-dependent synaptic plasticity. Neural Computation, 19 0 (6): 0 1468--1502, 2007. doi:10.1162/neco.2007.19.6.1468

    work page doi:10.1162/neco.2007.19.6.1468 2007

  17. [17]

    Neuromodulated spike-timing-dependent plasticity, and theory of three-factor learning rules

    Nicolas Fr \'e maux and Wulfram Gerstner. Neuromodulated spike-timing-dependent plasticity, and theory of three-factor learning rules. Frontiers in Neural Circuits, 9: 0 85, 2016. doi:10.3389/fncir.2015.00085

    work page doi:10.3389/fncir.2015.00085 2016

  18. [18]

    Furber, Francesco Galluppi, Steve Temple, and Luis A

    Steve B. Furber, Francesco Galluppi, Steve Temple, and Luis A. Plana. The spinnaker project. Proceedings of the IEEE, 102 0 (5): 0 652--665, 2014

    work page 2014

  19. [19]

    Drew, and L

    Stefano Fusi, Patrick J. Drew, and L. F. Abbott. Cascade models of synaptic plasticity: Toward the goal of understanding cerebral cortex. Neuron, 45 0 (4): 0 599--611, 2005

    work page 2005

  20. [20]

    Karunesh Ganguly and Jose M. Carmena. Emergence of a stable cortical map for neuroprosthetic control. PLoS Biology, 7 0 (7): 0 e1000153, 2009

    work page 2009

  21. [21]

    Triplet spike time-dependent plasticity in a floating-gate synapse

    Roshan Gopalakrishnan and Arindam Basu. Triplet spike time-dependent plasticity in a floating-gate synapse. IEEE Transactions on Neural Networks and Learning Systems, 28 0 (4): 0 778--790, 2017. doi:10.1109/TNNLS.2015.2506740

    work page doi:10.1109/tnnls.2015.2506740 2017

  22. [22]

    Meta-learning biologically plausible semi-supervised update rules

    Keren Gu, Sam Greydanus, Luke Metz, Niru Maheswaranathan, and Jascha Sohl-Dickstein. Meta-learning biologically plausible semi-supervised update rules. bioRxiv, pp.\ 2019.12.30.891184, 2019. doi:10.1101/2019.12.30.891184

    work page doi:10.1101/2019.12.30.891184 2019

  23. [23]

    Historical perspective and opportunity for computing in memory using floating-gate and resistive non-volatile computing including neuromorphic computing

    Jennifer Hasler and Arindam Basu. Historical perspective and opportunity for computing in memory using floating-gate and resistive non-volatile computing including neuromorphic computing. Neuromorphic Computing and Engineering, 5: 0 12001, 2025. doi:10.1088/2634-4386/ad9b4a

    work page doi:10.1088/2634-4386/ad9b4a 2025

  24. [24]

    SpQuant-SNN : Ultra-low precision membrane potential with sparse activations unlock the potential of on-device spiking neural networks applications

    Ahmed Hasssan, Jian Meng, Anupreetham Anupreetham, and Jae-sun Seo. SpQuant-SNN : Ultra-low precision membrane potential with sparse activations unlock the potential of on-device spiking neural networks applications. Frontiers in Neuroscience, 18: 0 1440000, 2024. doi:10.3389/fnins.2024.1440000

    work page doi:10.3389/fnins.2024.1440000 2024

  25. [25]

    Donald O. Hebb. The Organization of Behavior. Wiley, New York, 1949

    work page 1949

  26. [26]

    Steven Younger, and Peter R

    Sepp Hochreiter, A. Steven Younger, and Peter R. Conwell. Learning to learn using gradient descent. In Proceedings of the International Conference on Artificial Neural Networks (ICANN 2001), pp.\ 87--94, 2001

    work page 2001

  27. [27]

    Huang, Y

    X. Huang, Y. Xu, J. Hua, W. Yi, H. Yin, R. Hu, and S. Wang. A review on signal processing approaches to reduce calibration time in eeg-based brain--computer interface. Frontiers in Neuroscience, 15: 0 733546, 2021. doi:10.3389/fnins.2021.733546

    work page doi:10.3389/fnins.2021.733546 2021

  28. [28]

    Hardware-amenable structural learning for spike-based pattern classification using a simple model of active dendrites

    Shaista Hussain, Shih-Chii Liu, and Arindam Basu. Hardware-amenable structural learning for spike-based pattern classification using a simple model of active dendrites. Neural Computation, 27 0 (4): 0 845--897, 2015. doi:10.1162/NECO_a_00713

    work page doi:10.1162/neco_a_00713 2015

  29. [29]

    Memory and information processing in neuromorphic systems

    Giacomo Indiveri and Shih-Chii Liu. Memory and information processing in neuromorphic systems. Proceedings of the IEEE, 103 0 (8): 0 1379--1397, 2015

    work page 2015

  30. [30]

    Izhikevich

    Eugene M. Izhikevich. Solving the distal reward problem through linkage of STDP and dopamine signaling. Cerebral Cortex, 17 0 (10): 0 2443--2452, 2007

    work page 2007

  31. [31]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint, 2015

    work page 2015

  32. [32]

    Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, Demis Hassabis, Claudia Clopath, Dharshan Kumaran, and Raia Hadsell

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, et al. Overcoming catastrophic forgetting in neural networks. Proceedings of the National Academy of Sciences USA, 114 0 (13): 0 3521--3526, 2017. doi:10.1073/pnas.1611835114

    work page doi:10.1073/pnas.1611835114 2017

  33. [33]

    Learning with three factors: modulating hebbian plasticity with errors

    Lukasz Kusmierz, Takuya Isomura, and Taro Toyoizumi. Learning with three factors: modulating hebbian plasticity with errors. Current Opinion in Neurobiology, 46: 0 170--177, 2017. doi:10.1016/j.conb.2017.08.020

    work page doi:10.1016/j.conb.2017.08.020 2017

  34. [34]

    O'Doherty, Mikhail A

    Zhe Li, Joseph E. O'Doherty, Mikhail A. Lebedev, and Miguel A. L. Nicolelis. Adaptive decoding for brain--machine interfaces through bayesian parameter updates. Neural Computation, 23 0 (12): 0 3162--3204, 2011. doi:10.1162/NECO_a_00207

    work page doi:10.1162/neco_a_00207 2011

  35. [35]

    Lillicrap, Adam Santoro, Luke Marris, Colin J

    Timothy P. Lillicrap, Adam Santoro, Luke Marris, Colin J. Akerman, and Geoffrey Hinton. Backpropagation and the brain. Nature Reviews Neuroscience, 21 0 (6): 0 335--346, 2020. doi:10.1038/s41583-020-0275-8

    work page doi:10.1038/s41583-020-0275-8 2020

  36. [36]

    Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs

    Henry Markram, Joachim L "u bke, Michael Frotscher, and Bert Sakmann. Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs . Science, 275 0 (5297): 0 213--215, 1997

    work page 1997

  37. [37]

    Paxon Frady, and Friedrich T

    Szymon Mazurek, E. Paxon Frady, and Friedrich T. Sommer. Three-factor learning in spiking neural networks: An overview of methods and trends from a machine learning perspective. arXiv preprint, 2025

    work page 2025

  38. [38]

    Merolla, John V

    Paul A. Merolla, John V. Arthur, Rodrigo Alvarez-Icaza, et al. A million spiking-neuron integrated circuit with a scalable communication network and interface. Science, 345 0 (6197): 0 668--673, 2014

    work page 2014

  39. [39]

    Thomas Miconi, Jeff Clune, and Kenneth O. Stanley. Differentiable plasticity: Training plastic neural networks with backpropagation. In Proceedings of the 35th International Conference on Machine Learning (ICML 2018), volume 80 of Proceedings of Machine Learning Research, pp.\ 3559--3568, 2018

    work page 2018

  40. [40]

    First-spike based visual categorization using reward-modulated STDP

    Milad Mozafari et al. First-spike based visual categorization using reward-modulated STDP . arXiv preprint, 2018

    work page 2018

  41. [41]

    Neftci, Hesham Mostafa, and Friedemann Zenke

    Emre O. Neftci, Hesham Mostafa, and Friedemann Zenke. Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based optimization to spiking neural networks. IEEE Signal Processing Magazine, 36 0 (6): 0 51--63, 2019. doi:10.1109/MSP.2019.2931595

    work page doi:10.1109/msp.2019.2931595 2019

  42. [42]

    O'Doherty, Mariana M

    Joseph E. O'Doherty, Mariana M. B. Cardoso, Joseph G. Makin, and Philip N. Sabes. Nonhuman primate reaching with multichannel sensorimotor cortex electrophysiology. Dataset, 2017. URL https://doi.org/10.5281/zenodo.583331

    work page doi:10.5281/zenodo.583331 2017

  43. [43]

    Orsborn, Hamilton G

    Amy L. Orsborn, Hamilton G. Moorman, Simon A. Overduin, et al. Closed-loop decoder adaptation shapes neural plasticity for skillful neuroprosthetic control. Neuron, 82 0 (6): 0 1380--1393, 2014

    work page 2014

  44. [44]

    High performance communication by people with paralysis using an intracortical brain--computer interface

    Chethan Pandarinath et al. High performance communication by people with paralysis using an intracortical brain--computer interface. eLife, 6: 0 e18554, 2017. doi:10.7554/eLife.18554

    work page doi:10.7554/elife.18554 2017

  45. [45]

    Triplets of spikes in a model of spike timing-dependent plasticity

    Jean-Pascal Pfister and Wulfram Gerstner. Triplets of spikes in a model of spike timing-dependent plasticity. Journal of Neuroscience, 26 0 (38): 0 9673--9682, 2006

    work page 2006

  46. [46]

    Brian Premchand, Kyaw Kyar Toe, Chuanchu Wang, Shoeb Shaikh, Camilo Libedinsky, Kai Keng Ang, and Rosa Q. So. Decoding movement direction from cortical microelectrode recordings using an lstm-based neural network. In Proceedings of the 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pp.\ 3007--3010, Montr...

    work page doi:10.1109/embc44109.2020.9175593 2020

  47. [47]

    & Panda, P

    Kaushik Roy, Akhilesh Jaiswal, and Priyadarshini Panda. Towards spike-based machine intelligence with neuromorphic computing. Nature, 575 0 (7784): 0 607--617, 2019. doi:10.1038/s41586-019-1677-2

    work page doi:10.1038/s41586-019-1677-2 2019

  48. [48]

    Sadtler, Kevin M

    Patrick T. Sadtler, Kevin M. Quick, Matthew D. Golub, et al. Neural constraints on learning. Nature, 512 0 (7515): 0 423--426, 2014

    work page 2014

  49. [49]

    Meta-spikepropamine: learning to learn with synaptic plasticity in spiking neural networks

    Samuel Schmidgall, Joe Hays, and Alexander Ororbia. Meta-spikepropamine: learning to learn with synaptic plasticity in spiking neural networks. Frontiers in Neuroscience, 17: 0 1183321, 2023. doi:10.3389/fnins.2023.1183321

    work page doi:10.3389/fnins.2023.1183321 2023

  50. [50]

    Sebastian Seung

    H. Sebastian Seung. Learning in spiking neural networks by reinforcement of stochastic synaptic transmission. Neural Computation, 15 0 (8): 0 1821--1849, 2003

    work page 2003

  51. [51]

    Sunil Krishna, Jyotibdha Acharya, Arindam Basu, Kian Hsiang Chua, Nitish V

    Shoeb Shaikh, Na Duong Tran, G. Sunil Krishna, Jyotibdha Acharya, Arindam Basu, Kian Hsiang Chua, Nitish V. Thakor, and Haizhou Li. Neuromorphic decoders that outperform kalman filter and neural networks. IEEE Transactions on Biomedical Circuits and Systems, 13 0 (6): 0 1615--1624, 2019. doi:10.1109/TBCAS.2019.2944486

    work page doi:10.1109/tbcas.2019.2944486 2019

  52. [52]

    Meta-learning biologically plausible plasticity rules with random feedback pathways

    Navid Shervani-Tabar and Robert Rosenbaum. Meta-learning biologically plausible plasticity rules with random feedback pathways. Nature Communications, 14: 0 1555, 2023. doi:10.1038/s41467-023-37562-1

    work page doi:10.1038/s41467-023-37562-1 2023

  53. [53]

    Shih, Dean J

    Jerry J. Shih, Dean J. Krusienski, and Jonathan R. Wolpaw. Brain--computer interfaces in medicine. Mayo Clinic Proceedings, 87 0 (Suppl 1): 0 S107--S110, 2012

    work page 2012

  54. [54]

    S. Song, K. D. Miller, and L. F. Abbott. Competitive hebbian learning through spike-timing-dependent synaptic plasticity. Nature Neuroscience, 3 0 (9): 0 919--926, 2000

    work page 2000

  55. [55]

    Stavisky, Jonathan C

    David Sussillo, Sergey D. Stavisky, Jonathan C. Kao, Stephen I. Ryu, and Krishna V. Shenoy. Making brain--machine interfaces robust to future neural variability. Nature Communications, 7: 0 13749, 2016. doi:10.1038/ncomms13749

    work page doi:10.1038/ncomms13749 2016

  56. [56]

    Taeckens and Sahil Shah

    Elijah A. Taeckens and Sahil Shah. A spiking neural network with continuous local learning for robust online brain--machine interface. Journal of Neural Engineering, 20 0 (6): 0 066042, 2024. doi:10.1088/1741-2552/ad1787

    work page doi:10.1088/1741-2552/ad1787 2024

  57. [57]

    Turrigiano and Sacha B

    Gina G. Turrigiano and Sacha B. Nelson. Homeostatic plasticity in the developing nervous system. Nature Reviews Neuroscience, 5 0 (2): 0 97--107, 2004. doi:10.1038/nrn1328

    work page doi:10.1038/nrn1328 2004

  58. [58]

    Wolpaw, Niels Birbaumer, Dennis J

    Jonathan R. Wolpaw, Niels Birbaumer, Dennis J. McFarland, Gert Pfurtscheller, and Theresa M. Vaughan. Brain--computer interfaces for communication and control. Clinical Neurophysiology, 113 0 (6): 0 767--791, 2002

    work page 2002

  59. [59]

    Black, David Mumford, Yun Gao, Elie Bienenstock, and John P

    Wei Wu, Michael J. Black, David Mumford, Yun Gao, Elie Bienenstock, and John P. Donoghue. Modeling and decoding motor cortical activity using a switching kalman filter. IEEE Transactions on Biomedical Engineering, 51 0 (6): 0 933--942, 2004. doi:10.1109/TBME.2004.826666

    work page doi:10.1109/tbme.2004.826666 2004

  60. [60]

    SuperSpike: Supervised Learning in Multilayer Spiking Neural Networks

    Friedemann Zenke and Surya Ganguli. Superspike: Supervised learning in multilayer spiking neural networks. Neural Computation, 30 0 (6): 0 1514--1541, 2018. doi:10.1162/neco_a_01086

    work page doi:10.1162/neco_a_01086 2018

  61. [61]

    Friedemann Zenke and Tim P. Vogels. The remarkable robustness of surrogate gradient learning for instilling complex function in spiking neural networks. Neural Computation, 33 0 (4): 0 899--925, 2021. doi:10.1162/neco_a_01367

    work page doi:10.1162/neco_a_01367 2021

  62. [62]

    Continual learning through synaptic intelligence

    Friedemann Zenke, Ben Poole, and Surya Ganguli. Continual learning through synaptic intelligence. In Proceedings of the 34th International Conference on Machine Learning (ICML 2017), volume 70 of Proceedings of Machine Learning Research, pp.\ 3987--3995, 2017

    work page 2017

  63. [63]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...

    work page

  64. [64]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...

    work page

  65. [65]

    FABulous

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibsetup #1 @NAT@ctr @ @openbib .11em \@plus.33em \@minus.07em 4000 4000 `\.\@m @bibit...

    work page doi:10.1109/iscas56072.2025.11043277 2025