pith. sign in

arxiv: 2503.02642 · v3 · submitted 2025-03-04 · 🧬 q-bio.NC · cs.ET· cs.LG· cs.NE

Spike-based alignment learning solves the weight transport problem

Timo Gierlich , Andreas Baumbach , Akos F. Kungl , Kevin Max , Mihai A. Petrovici This is my paper

Pith reviewed 2026-05-23 01:58 UTC · model grok-4.3

classification 🧬 q-bio.NC cs.ETcs.LGcs.NE
keywords spike-based alignment learningweight transport problemspiking neural networksHebbian plasticityanti-Hebbian plasticitylocal learning ruleserror backpropagationneuromorphic computing
0
0 comments X

The pith

Spike timing statistics allow synapses to recover the true local gradient by correcting asymmetry in reciprocal connections.

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

The paper introduces spike-based alignment learning (SAL) to solve the weight transport problem that prevents local gradient computation in spiking networks. SAL extracts and corrects asymmetries between effective reciprocal connections using only local spike timing statistics. This occurs through an interplay of Hebbian and anti-Hebbian plasticity, without requiring symmetric weights or global signals. The rule benefits from noise and mitigates neuron and synapse variability. It is demonstrated to improve convergence in probabilistic networks, align feedback in cortical hierarchies, and support competitive deep network performance with fully local plasticity.

Core claim

SAL uses spike timing statistics to extract and correct the asymmetry between effective reciprocal connections; synapses thereby recover the true local gradient via an interplay between Hebbian and anti-Hebbian plasticity. Apart from being spike-based and fully local, the mechanism takes advantage of noise and alleviates discrepancies from neuron and synapse variability. Demonstrations show improved convergence in probabilistic spiking networks, effective alignment of feedback weights in cortical microcircuit hierarchies for correct error backpropagation, and competitive performance in deep networks using only local plasticity.

What carries the argument

Spike-based alignment learning (SAL), a complementary learning rule that aligns effective reciprocal connections using spike timing statistics through Hebbian and anti-Hebbian plasticity.

If this is right

  • SAL significantly improves convergence to the target distribution in probabilistic spiking networks versus Hebbian plasticity alone.
  • In neuronal hierarchies based on cortical microcircuits, SAL aligns feedback weights to the forward pathway and enables correct backpropagation of feedback errors.
  • SAL enables competitive performance in deep networks using only local plasticity for weight transport.
  • The approach alleviates discrepancies arising from neuron and synapse variability in physical networks.

Where Pith is reading between the lines

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

  • Biological learning could implement gradient descent using only local timing-based rules without assuming weight symmetry.
  • Neuromorphic hardware could apply SAL to achieve backpropagation-like updates with strictly local computation.
  • The timing-based alignment may generalize to non-spiking networks or other plasticity mechanisms if similar statistics are available.

Load-bearing premise

Spike timing statistics alone suffice to recover an accurate local gradient across varied network architectures and noise levels without additional global signals or symmetric weights.

What would settle it

A controlled simulation where spike timing is randomized or uncorrelated with connection asymmetries while other variables are held fixed, showing whether SAL still produces alignment and improved learning.

Figures

Figures reproduced from arXiv: 2503.02642 by Akos F. Kungl, Andreas Baumbach, Kevin Max, Mihai A. Petrovici, Timo Gierlich.

Figure 1
Figure 1. Figure 1: The weight transport problem in physical neu [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Principles of spike-based alignment learning. a) [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Working principle of SSNs. a) An SSN consists of a recurrent spiking network with symmetric reciprocal connec￾tions in theory, but asymmetric ones in practice. b) Each neu￾ron fires stochastically as a function of the membrane potential (blue), which gives rise to a sampling process from an underlying distribution p(z). The refractory state of each neuron is mapped to a binary variable z ∈ [0, 1] (green an… view at source ↗
Figure 4
Figure 4. Figure 4: Synaptic noise in SSNs. An SSN is initialized with a symmetric weight matrix and trained with pure STDP-based wake-sleep (blue curve), which serves as a baseline (no noise). We also train a noised weight matrix (additive Gaussian noise with standard deviation σ noise syn ) with (green) and without (orange) a SAL phase in addition to STDP-based wake-sleep learning and the Kolen-Pollack algorithm (KP, red). … view at source ↗
Figure 5
Figure 5. Figure 5: Plasticity noise in SSNs. The basic configuration of [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: SAL enables accurate BP in a spiking cortical microcircuit model. a) [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of SAL with other symmetrization schemes in a deep network. a) [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Time evolution of SAL in the SALNet for the three layers with constant forward weights for different learning rates. The learning rate serves as the crucial tuning parameter in the trade-off between fast symmetrization and small alignment angles. The upper x-axis displays the number of training epochs in [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Symmetrization and sign changes in networks [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: SAL works with different PSP shapes. a) We consider a two-neuron system together with three common PSP kernels. b) STDDs at W12 = W21 = 1 for three different bias combinations. The STDDs of the α−kernels show a slight asymmetry if b1 ̸= b2. c) Phase plane diagram of reciprocal weights for nine bias combinations. For clarity, only the attractors are shown. The biases used here are marked as squares in the … view at source ↗
Figure 11
Figure 11. Figure 11: a) STDD deviation δ(t) for Wij = Wji + ε (with ε = 0.005) and rectangular PSPs, for positive (left) and negative weights (right). b) Integrated deviation ∆(t) for positive (left) and negative (right) weights. Wij = Wji, we have p + 0 = ˆp − 0 , so the average weight update ∆W˙ ij = 0. Thus, Wij = Wji is a unique fixed point. We now show that the fixed points are stable under small perturbations of one wei… view at source ↗
Figure 12
Figure 12. Figure 12: Schematic of the microcircuit model: bio-plausible transportation of error signals and local error representation in apical dendrites. Adapted from [51] 4.5 Deep symmetrization network (SymmNet) The network architecture of the SymmNet is identical to the design in [54]. It consists of two different but linked networks: a classical (non-spiking) artificial neural network (ANN) for the training of the forwa… view at source ↗
Figure 13
Figure 13. Figure 13: Comparison between the analytical method [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Choosing the optimal Kolen-Pollack weight decay rate [PITH_FULL_IMAGE:figures/full_fig_p024_14.png] view at source ↗
read the original abstract

In both machine learning and in computational neuroscience, plasticity in functional neural networks is frequently expressed as gradient descent on a cost. Often, this imposes symmetry constraints that are difficult to reconcile with local computation, as is required for biological networks or neuromorphic hardware. For example, wake-sleep learning in networks characterized by Boltzmann distributions assumes symmetric connectivity. Similarly, the error backpropagation algorithm is notoriously plagued by the weight transport problem between the representation and the error stream. Existing solutions such as feedback alignment circumvent the problem by deferring to the robustness of these algorithms to weight asymmetry. However, they scale poorly with network size and depth. We introduce spike-based alignment learning (SAL), a complementary learning rule for spiking neural networks, which uses spike timing statistics to extract and correct the asymmetry between effective reciprocal connections. Apart from being spike-based and fully local, our proposed mechanism takes advantage of noise. Based on an interplay between Hebbian and anti-Hebbian plasticity, synapses can thereby recover the true local gradient. This also alleviates discrepancies that arise from neuron and synapse variability -- an omnipresent property of physical neuronal networks. We demonstrate the efficacy of our mechanism using different spiking network models. First, SAL can significantly improve convergence to the target distribution in probabilistic spiking networks versus Hebbian plasticity alone. Second, in neuronal hierarchies based on cortical microcircuits, SAL effectively aligns feedback weights to the forward pathway, thus allowing the backpropagation of correct feedback errors. Third, our approach enables competitive performance in deep networks using only local plasticity for weight transport.

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

0 major / 3 minor

Summary. The manuscript introduces spike-based alignment learning (SAL), a complementary local plasticity rule for spiking neural networks. SAL uses spike timing statistics to extract and correct asymmetries between effective reciprocal connections via an interplay of Hebbian and anti-Hebbian plasticity, thereby recovering the true local gradient without requiring symmetric weights or global signals. The mechanism is shown to exploit rather than suffer from noise and variability. Efficacy is demonstrated in three regimes: improved convergence to target distributions in probabilistic spiking networks, alignment of feedback weights in cortical microcircuit hierarchies to enable correct error backpropagation, and competitive performance in deep networks using only local plasticity for weight transport.

Significance. If the quantitative results hold, SAL provides a fully local, spike-based solution to the weight transport problem that is compatible with biological constraints and neuromorphic hardware. The approach is notable for treating noise and neuronal variability as exploitable features rather than obstacles, and for its potential to bridge gradient-based learning in machine learning with local computation in neuroscience. The three demonstration settings offer a broad test of the mechanism across network types.

minor comments (3)
  1. Abstract: the claims of 'significantly improve convergence' and 'competitive performance' would be strengthened by inclusion of at least one quantitative metric (e.g., KL divergence reduction or test accuracy) with error bars or statistical comparison to baselines.
  2. The manuscript would benefit from an explicit statement of the precise conditions under which the Hebbian/anti-Hebbian interplay is guaranteed to recover the gradient (e.g., in terms of firing rate regimes or correlation assumptions).
  3. Figure captions and methods should clarify whether the reported alignments are measured by cosine similarity, weight correlation, or another metric, and whether results are averaged over multiple random seeds.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the constructive and positive review, including the recommendation for minor revision. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces SAL as a local, spike-based mechanism relying on Hebbian/anti-Hebbian interplay to correct weight asymmetry using timing statistics. No equations, fitted parameters, or self-citations are shown in the provided text that reduce the claimed gradient recovery to a definition or input by construction. The three demonstration regimes are presented as empirical validation of an independent local process rather than tautological predictions. The central claim does not invoke uniqueness theorems or ansatzes from prior self-work in a load-bearing way visible here. This is the expected non-finding for a mechanism paper whose derivation chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the mechanism is described at the level of plasticity rules without mathematical specification.

pith-pipeline@v0.9.0 · 5830 in / 1156 out tokens · 35671 ms · 2026-05-23T01:58:42.607888+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

121 extracted references · 121 canonical work pages · 5 internal anchors

  1. [1]

    Hopfield, J. J. Neural Networks and Physical Systems with Emergent Collective Computational Abilities. Proceedings of the National Academy of Sciences 79, 2554–2558 (Apr. 1982)

    work page 1982

  2. [2]

    Hinton, G. E. & Sejnowski, T. J. Optimal perceptual infer- ence in Proceedings of the IEEE conference on Computer Vision and Pattern Recognition 448 (1983), 448–453

    work page 1983

  3. [3]

    H., Hinton, G

    Ackley, D. H., Hinton, G. E. & Sejnowski, T. J. A Learning Algorithm for Boltzmann Machines*. Cognitive Science 9, 147–169. issn: 0364-0213, 1551-6709 (Jan. 1985)

    work page 1985

  4. [4]

    Hinton, G. E. Training products of experts by minimizing contrastive divergence. Neural computation 14, 1771–1800 (2002)

    work page 2002

  5. [5]

    Taylor Expansion of the Accumulated Rounding Error

    Linnainmaa, S. Taylor Expansion of the Accumulated Rounding Error. BIT Numerical Mathematics 16, 146–

    work page

  6. [6]

    issn: 1572-9125 (June 1976)

    work page 1976

  7. [7]

    Werbos, P. J. in System Modeling and Optimization 762– 770 (Springer-Verlag, Sept. 2005). http://dx.doi.org/ 10.1007/BFb0006203

    work page doi:10.1007/bfb0006203 2005

  8. [8]

    E., Hinton, G

    Rumelhart, D. E., Hinton, G. E. & Williams, R. J. Learn- ing Representations by Back-Propagating Errors. Nature 323, 533–536. issn: 1476-4687 (Oct. 1986)

    work page 1986

  9. [9]

    Whittington, J. C. R. & Bogacz, R. Theories of Error Back- Propagation in the Brain. Trends in Cognitive Sciences 23, 235–250. issn: 1364-6613, 1879-307X (Mar. 2019)

    work page 2019

  10. [10]

    Richards, B. A. et al. A Deep Learning Framework for Neuroscience. Nature Neuroscience 22, 1761–1770. issn: 1546-1726 (Nov. 2019)

    work page 2019

  11. [11]

    P., Santoro, A., Marris, L., Akerman, C

    Lillicrap, T. P., Santoro, A., Marris, L., Akerman, C. J. & Hinton, G. Backpropagation and the Brain. Nature Re- views Neuroscience 21, 335–346. issn: 1471-0048 (Apr. 2020)

    work page 2020

  12. [12]

    & Goaillard, J.-M

    Marder, E. & Goaillard, J.-M. Variability, Compensation and Homeostasis in Neuron and Network Function.Nature Reviews Neuroscience 7, 563–574 (2006)

    work page 2006

  13. [13]

    Watt, A. J. & Desai, N. S. Homeostatic plasticity and STDP: keeping a neuron’s cool in a fluctuating world. Frontiers in synaptic neuroscience 2, 1486 (2010)

    work page 2010

  14. [14]

    Abbott, L. F. & Nelson, S. B. Synaptic plasticity: taming the beast. Nature neuroscience 3, 1178–1183. issn: 1546-

    work page

  15. [15]

    http://dx.doi.org/10.1038/81453 (Nov. 2000)

    work page doi:10.1038/81453 2000

  16. [16]

    Davis, G. W. & Bezprozvanny, I. Maintaining the stabil- ity of neural function: a homeostatic hypothesis. Annual review of physiology 63, 847–869 (2001). 24

    work page 2001

  17. [17]

    & Kirkwood, A

    Lee, H.-K. & Kirkwood, A. Mechanisms of homeostatic synaptic plasticity in vivo. Frontiers in Cellular Neuro- science 13, 520 (2019)

    work page 2019

  18. [18]

    & Prescott, S

    Yang, J. & Prescott, S. A. Homeostatic regulation of neu- ronal function: importance of degeneracy and pleiotropy. Frontiers in Cellular Neuroscience 17, 1184563 (2023)

    work page 2023

  19. [19]

    Bartol Jr, T. M. et al. Nanoconnectomic upper bound on the variability of synaptic plasticity.elife 4, e10778 (2015)

    work page 2015

  20. [20]

    Noguchi, J. et al. Bidirectional in vivo structural den- dritic spine plasticity revealed by two-photon glutamate uncaging in the mouse neocortex. Scientific reports 9, 13922 (2019)

    work page 2019

  21. [21]

    D¨ urst, C. D. et al. Vesicular release probability sets the strength of individual Schaffer collateral synapses. Nature Communications 13, 6126 (2022)

    work page 2022

  22. [22]

    E., Tigaret, C

    Rodrigues, Y. E., Tigaret, C. M., Marie, H., O’Donnell, C. & Veltz, R. A stochastic model of hippocampal synaptic plasticity with geometrical readout of enzyme dynamics. eLife 12. issn: 2050-084X. http://dx.doi.org/10.7554/ eLife.80152 (Aug. 2023)

    work page 2050

  23. [23]

    Billaudelle, S. et al. Versatile Emulation of Spiking Neu- ral Networks on an Accelerated Neuromorphic Substrate in 2020 IEEE International Symposium on Circuits and Systems (ISCAS) (Oct. 2020), 1–5

    work page 2020

  24. [24]

    Atoui, A. et al. Multi-timescale synaptic plasticity on ana- log neuromorphic hardware in 2025 Neuro Inspired Com- putational Elements (NICE) (2025), 1–9

    work page 2025

  25. [25]

    & Maass, W

    Buesing, L., Bill, J., Nessler, B. & Maass, W. Neural dy- namics as sampling: a model for stochastic computation in recurrent networks of spiking neurons. PLoS computa- tional biology 7, e1002211 (2011)

    work page 2011

  26. [26]

    A., Bill, J., Bytschok, I., Schemmel, J

    Petrovici, M. A., Bill, J., Bytschok, I., Schemmel, J. & Meier, K. Stochastic inference with spiking neurons in the high-conductance state. Physical Review E 94, 042312 (2016)

    work page 2016

  27. [27]

    & Fiser, J

    Berkes, P., Orb´ an, G., Lengyel, M. & Fiser, J. Spontaneous cortical activity reveals hallmarks of an optimal internal model of the environment. Science 331, 83–87 (2011)

    work page 2011

  28. [28]

    M., Berkes, P

    Haefner, R. M., Berkes, P. & Fiser, J. Perceptual decision- making as probabilistic inference by neural sampling. Neu- ron 90, 649–660 (2016)

    work page 2016

  29. [29]

    & Lengyel, M

    Orb´ an, G., Berkes, P., Fiser, J. & Lengyel, M. Neural vari- ability and sampling-based probabilistic representations in the visual cortex. Neuron 92, 530–543 (2016)

    work page 2016

  30. [30]

    Kungl, A. F. et al. Accelerated physical emulation of bayesian inference in spiking neural networks. Frontiers in neuroscience 13, 1201 (2019)

    work page 2019

  31. [31]

    Czischek, S. et al. Spiking Neuromorphic Chip Learns En- tangled Quantum States. SciPost Physics 12, 039. issn: 2542-4653 (Jan. 2022)

    work page 2022

  32. [32]

    Klassert, R., Baumbach, A., Petrovici, M. A. & G¨ art- tner, M. Variational Learning of Quantum Ground States on Spiking Neuromorphic Hardware. iScience 25, 104707. issn: 2589-0042 (Aug. 2022)

    work page 2022

  33. [33]

    & Cauwenberghs, G

    Neftci, E., Das, S., Pedroni, B., Kreutz-Delgado, K. & Cauwenberghs, G. Event-driven contrastive divergence for spiking neuromorphic systems. Frontiers in neuroscience 7, 272 (2014)

    work page 2014

  34. [34]

    The Recent Excitement about Neural Networks

    Crick, F. The Recent Excitement about Neural Networks. Nature 337, 129–132. issn: 1476-4687 (Jan. 1989)

    work page 1989

  35. [35]

    J., Prakash, P

    Lansdell, B. J., Prakash, P. R. & Kording, K. P. Learn- ing to solve the credit assignment problem. arXiv preprint arXiv:1906.00889 (2019)

    work page arXiv 1906

  36. [36]

    Ernoult, M. M. et al. Towards Scaling Difference Target Propagation by Learning Backprop Targets in Proceedings of the 39th International Conference on Machine Learning (PMLR, June 2022), 5968–5987

    work page 2022

  37. [37]

    & Seung, H

    Xie, X. & Seung, H. S. Equivalence of Backpropagation and Contrastive Hebbian Learning in a Layered Network. Neural Computation 15, 441–454. issn: 0899-7667 (Feb. 2003)

    work page 2003

  38. [38]

    & Bengio, Y

    Scellier, B. & Bengio, Y. Equilibrium Propagation: Bridg- ing the Gap between Energy-Based Models and Backprop- agation. Frontiers in Computational Neuroscience 11. issn: 1662-5188. http : / / dx . doi . org / 10 . 3389 / fncom . 2017.00024 (May 2017)

    work page arXiv 2017

  39. [39]

    & Senn, W

    Sacramento, J., Ponte Costa, R., Bengio, Y. & Senn, W. Dendritic Cortical Microcircuits Approximate the Back- propagation Algorithm in Advances in Neural Informa- tion Processing Systems (eds Bengio, S. et al. ) 31 (Cur- ran Associates, Inc., 2018). https : / / proceedings . neurips . cc / paper _ files / paper / 2018 / file / 1dc3a89d0d440ba31729b0ba74b9...

    work page 2018

  40. [40]

    & Roelfsema, P

    Pozzi, I., Bohte, S. & Roelfsema, P. Attention-Gated Brain Propagation: How the brain can implement reward-based error backpropagation. Advances in neural information processing systems 33, 2516–2526 (2020)

    work page 2020

  41. [41]

    Kolen, J. F. & Pollack, J. B. Backpropagation without Weight Transport in Proceedings of 1994 IEEE Interna- tional Conference on Neural Networks (ICNN’94) 3 (June 1994), 1375–1380

    work page 1994

  42. [42]

    Payeur, A., Guerguiev, J., Zenke, F., Richards, B. A. & Naud, R. Burst-Dependent Synaptic Plasticity Can Co- ordinate Learning in Hierarchical Circuits. Nature Neuro- science 24, 1010–1019. issn: 1546-1726 (July 2021)

    work page 2021

  43. [43]

    Roelfsema, P. R. & Ooyen, A. v. Attention-gated reinforce- ment learning of internal representations for classification. Neural computation 17, 2176–2214 (2005)

    work page 2005

  44. [44]

    A Biologically Plausible Learning Rule for Deep Learning in the Brain

    Pozzi, I., Boht´ e, S. & Roelfsema, P. A biologically plausible learning rule for deep learning in the brain. arXiv preprint arXiv:1811.01768 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  45. [45]

    Haider, P. et al. Latent Equilibrium: A Unified Learning Theory for Arbitrarily Fast Computation with Arbitrarily Slow Neurons in Advances in Neural Information Process- ing Systems 34 (Curran Associates, Inc., 2021), 17839– 17851

    work page 2021

  46. [46]

    Ellenberger, B. et al. Backpropagation through space, time and the brain. Nature Communications 17. issn: 2041-

    work page 2041

  47. [47]

    http://dx.doi.org/10.1038/s41467-025-66666-z (Dec. 2025)

    work page doi:10.1038/s41467-025-66666-z 2025

  48. [48]

    Senn, W. et al. A neuronal least-action principle for real- time learning in cortical circuits. ELife 12, RP89674 (2024)

    work page 2024

  49. [49]

    & Bengio, Y

    Lee, D.-H., Zhang, S., Fischer, A. & Bengio, Y. Differ- ence target propagation in Joint european conference on machine learning and knowledge discovery in databases (Springer International Publishing, 2015), 498–515. http: //dx.doi.org/10.1007/978-3-319-23528-8_31

    work page doi:10.1007/978-3-319-23528-8_31 2015

  50. [50]

    & Tweed, D

    Akrout, M., Wilson, C., Humphreys, P., Lillicrap, T. & Tweed, D. B. Deep Learning without Weight Transport in Advances in Neural Information Processing Systems 32 (Curran Associates, Inc., 2019)

    work page 2019

  51. [51]

    P., Cownden, D., Tweed, D

    Lillicrap, T. P., Cownden, D., Tweed, D. B. & Akerman, C. J. Random Synaptic Feedback Weights Support Error Backpropagation for Deep Learning. Nature Communica- tions 7, 13276. issn: 2041-1723 (Dec. 2016)

    work page 2041

  52. [52]

    Bartunov, S. et al. Assessing the Scalability of Biologically- Motivated Deep Learning Algorithms and Architectures in Advances in Neural Information Processing Sys- tems 31 (Curran Associates, Inc., 2018). https : / / proceedings . neurips . cc / paper / 2018 / hash / 63c3ddcc7b23daa1e42dc41f9a44a873-Abstract.html

    work page 2018

  53. [53]

    Feedback alignment in deep convolutional networks

    Moskovitz, T. H., Litwin-Kumar, A. & Abbott, L. F. Feedback alignment in deep convolutional networks.arXiv preprint arXiv:1812.06488 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  54. [54]

    Max, K. et al. Learning Efficient Backprojections across Cortical Hierarchies in Real Time. Nature Machine Intel- ligence 6, 619–630. issn: 2522-5839 (June 2024). 25

    work page 2024

  55. [55]

    Burbank, K. S. Mirrored STDP implements autoencoder learning in a network of spiking neurons. PLoS computa- tional biology 11, e1004566 (2015)

    work page 2015

  56. [56]

    Guerguiev, J., Lillicrap, T. P. & Richards, B. A. Towards Deep Learning with Segregated Dendrites. eLife 6 (ed Latham, P.) e22901. issn: 2050-084X (Dec. 2017)

    work page 2050

  57. [57]

    Guerguiev, J., Kording, K. P. & Richards, B. A. Spike-Based Causal Inference for Weight Alignment. arXiv:1910.01689. arXiv: 1910.01689 (2019)

    work page arXiv 1910

  58. [58]

    & van Gerven, M

    Ahmad, N., Ambrogioni, L. & van Gerven, M. A. Overcoming the weight transport problem via spike- timing-dependent weight inference. arXiv preprint arXiv:2003.03988 (2020)

    work page arXiv 2003

  59. [59]

    Petrovici, M. A. et al. Characterization and Compensation of Network-Level Anomalies in Mixed-Signal Neuromor- phic Modeling Platforms. PLOS ONE 9, e108590. issn: 1932-6203 (Oct. 2014)

    work page 1932

  60. [60]

    Pehle, C. et al. The BrainScaleS-2 accelerated neuromor- phic system with hybrid plasticity. Frontiers in Neuro- science 16, 795876 (2022)

    work page 2022

  61. [61]

    & Schemmel, J

    Gr¨ ubl, A., Billaudelle, S., Cramer, B., Karasenko, V. & Schemmel, J. Verification and Design Methods for the BrainScaleS Neuromorphic Hardware System. Journal of Signal Processing Systems 92, 1277–1292. issn: 1939-8115. http://dx.doi.org/10.1007/s11265-020-01558-7 (July 2020)

    work page doi:10.1007/s11265-020-01558-7 1939

  62. [62]

    Groupings of nonpyramidal and pyramidal cells with specific physiological and morphological charac- teristics in rat frontal cortex

    Kawaguchi, Y. Groupings of nonpyramidal and pyramidal cells with specific physiological and morphological charac- teristics in rat frontal cortex. Journal of neurophysiology 69, 416–431 (1993)

    work page 1993

  63. [63]

    Liu, Y. et al. Neuronal diversity and stereotypy at multiple scales through whole brain morphometry. Nature Commu- nications 15, 10269 (2024)

    work page 2024

  64. [64]

    S., Rogal, A

    Mednikova, Y. S., Rogal, A. V., et al. Heterogeneity of the Neural Composition of Cortical Regions as a Condition for a Wide Range Regulating Spontaneous Activity. Journal of Behavioral and Brain Science 10, 220 (2020)

    work page 2020

  65. [65]

    Rathour, R. K. & Kaphzan, H. Voltage-gated ion channels and the variability in information transfer. Frontiers in Cellular Neuroscience 16, 906313 (2022)

    work page 2022

  66. [66]

    A., Selen, L

    Faisal, A. A., Selen, L. P. & Wolpert, D. M. Noise in the nervous system. Nature reviews neuroscience 9, 292–303 (2008)

    work page 2008

  67. [67]

    Dynamics of sparsely connected networks of ex- citatory and inhibitory spiking neurons

    Brunel, N. Dynamics of sparsely connected networks of ex- citatory and inhibitory spiking neurons. Journal of com- putational neuroscience 8, 183–208 (2000)

    work page 2000

  68. [68]

    & Brunel, N

    Fourcaud, N. & Brunel, N. Dynamics of the firing probabil- ity of noisy integrate-and-fire neurons.Neural computation 14, 2057–2110 (2002)

    work page 2057

  69. [69]

    Jordan, J. et al. Deterministic networks for probabilistic computing. Scientific reports 9, 18303 (2019)

    work page 2019

  70. [70]

    Dold, D. et al. Stochasticity from Function — Why the Bayesian Brain May Need No Noise.Neural Networks 119, 200–213. issn: 0893-6080 (Nov. 2019)

    work page 2019

  71. [71]

    M¨ uller, E. et al. Extending BrainScaleS OS for BrainScaleS-2 tech. rep. (Heidelberg, Germany, Mar. 2020). arXiv: 2003.13750 [cs.NE]

    work page arXiv 2020

  72. [72]

    & Veniero, D

    Koch, G., Ponzo, V., Di Lorenzo, F., Caltagirone, C. & Veniero, D. Hebbian and anti-Hebbian spike-timing- dependent plasticity of human cortico-cortical connections. Journal of Neuroscience 33, 9725–9733 (2013)

    work page 2013

  73. [73]

    Pfeil, T. et al. Six Networks on a Universal Neuromor- phic Computing Substrate. Frontiers in neuroscience 7, 11 (2013)

    work page 2013

  74. [74]

    & Linares-Barranco, B

    Serrano-Gotarredona, T., Masquelier, T., Prodromakis, T., Indiveri, G. & Linares-Barranco, B. STDP and STDP vari- ations with memristors for spiking neuromorphic learning systems. Frontiers in neuroscience 7, 2 (2013)

    work page 2013

  75. [75]

    Qiao, N. et al. A reconfigurable on-line learning spiking neuromorphic processor comprising 256 neurons and 128K synapses. Frontiers in neuroscience 9, 141 (2015)

    work page 2015

  76. [76]

    Davies, M. et al. Loihi: A neuromorphic manycore proces- sor with on-chip learning. IEEE Micro 38, 82–99. issn: 1937-4143. http : / / dx . doi . org / 10 . 1109 / MM . 2018 . 112130359 (Jan. 2018)

    work page 1937

  77. [77]

    M., Ma, W

    Pouget, A., Beck, J. M., Ma, W. J. & Latham, P. E. Proba- bilistic brains: knowns and unknowns.Nature neuroscience 16, 1170–1178 (2013)

    work page 2013

  78. [78]

    & Lengyel, M

    Fiser, J., Berkes, P., Orb´ an, G. & Lengyel, M. Statisti- cally Optimal Perception and Learning: From Behavior to Neural Representations. Trends in Cognitive Sciences 14, 119–130. issn: 1364-6613 (Mar. 2010)

    work page 2010

  79. [79]

    E., et al

    Krizhevsky, A., Hinton, G. E., et al. Learning Multiple Layers of Features from Tiny Images (2009)

    work page 2009

  80. [80]

    Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms

    Xiao, H., Rasul, K. & Vollgraf, R. Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learn- ing Algorithms 2017. arXiv: 1708.07747 [cs.LG] . https: //arxiv.org/abs/1708.07747

    work page internal anchor Pith review Pith/arXiv arXiv 2017

Showing first 80 references.