Pith. sign in

REVIEW 3 major objections 4 minor 47 references

Storing overlapping associative memories on latent manifolds in low-rank spiking networks

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper shows that low-rank spiking networks with purely inhibitory connectivity can store up to about one overlapping memory per two neurons as stable attractors, by placing memories at the vertices of a hypercubic latent manifold.

desk verdict A clean construction for overlapping spiking associative memory with linear empirical capacity, but the theory is weaker than it looks and several implementation details need tightening. read the letter →

arxiv 2411.17485 v2 pith:PDHCHBD2 submitted 2024-11-26 q-bio.NC cs.LGcs.NE

William F. Podlaski , Christian K. Machens This is my paper · ORCID
classification q-bio.NCcs.LGcs.NE MSC 92B20
keywords associativememoryspikingneuralnetworkslow-rankconnectivityHopfieldnetworklatentmanifoldstoragecapacitypatterncompletionall-inhibitory
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Associative memory in spiking neural networks has usually been limited to a few non-overlapping patterns. This paper claims that an all-inhibitory, rank-constrained spiking network can store many overlapping binary memories by arranging them at the vertices of a hypercube in a low-dimensional latent space. With pseudoinverse or optimized decoders, the authors report that storage capacity grows linearly with neuron number—roughly $0.3N$ and $0.5N$ memories respectively—and that noisy initial states reliably complete to the stored pattern. If correct, this gives a concrete recipe for high-capacity, content-addressable memory in biologically plausible circuits, and a direct bridge between spiking dynamics and the classical Hopfield model.

What carries the argument

The load-bearing object is the rank-$(K+1)$ convex, piecewise-linear latent manifold: recurrent weights are decomposed as $W=ED$, the latent variable is $y=Dr$, and each neuron's threshold becomes a hyperplane face of a hypercubic cone. The decoders are what implement memory: requiring $D\eta_\mu=\xi_\mu$ makes the leak dynamics at each vertex point back toward the vertex, turning it into an attractor. The extra latent dimension adds a constant inhibitory component so the boundary dynamics are stable, and a negative self-connection $-\gamma$ (tuned per load) prevents each neuron from stabilizing only its own face.

What would settle it

Run the optimized rule at load $p/N=0.55$ in a network of $N=400$ neurons and measure the final overlap for each stored pattern after recall; if any pattern's overlap falls below the recall criterion, the reported $0.5N$ capacity is not a sharp limit. A second test is to sample each active neuron's target rate independently around $\kappa$; if the attractors lose stability at small spreads near capacity, the equal-rate assumption is doing the work.

Watch

Extended reading notes

Core claim

The central result is that each memory is a stable attractor exactly when the decoder satisfies $D\eta_\mu=\xi_\mu$ for every pattern $\mu$, with neural patterns defined as $\eta_\mu=\mathrm{ReLU}(\kappa E\xi_\mu)$. With the encoding $E=[I_K;-I_K]$, the $N=2K$ neurons form the faces of a $K$-dimensional hypercube (plus one extra latent dimension that guarantees all-inhibitory stability), and the $2^K$ vertices are the candidate binary patterns. The Hebbian rule cannot meet the equality constraint and produces spurious mixtures; the pseudoinverse rule $D=\xi Q^{-1}\xi^\top E^\top$ stores up to about $0.3N$ patterns, and the optimized least-squares rule extends this to about $0.5N$ in simulations of up to 1600 neurons. These networks also complete patterns from noisy initial states, correcting up to 30% flipped bits at low load, and the same construction stores correlated MNIST-like patterns.

Load-bearing premise

The construction assumes that active neurons in a memory fire at one common average rate and that the latent-manifold equations capture the spiking network exactly, so that stabilizing each vertex in the reduced system is enough to stabilize the spiking attractor.

Editorial extensions

If this is right

  • The pseudoinverse and optimized decoders turn hypercube vertices into stable spiking attractors, while the Hebbian rule fails because it cannot satisfy the equality constraint $D\eta_\mu=\xi_\mu$.
  • Storage capacity scales linearly with the number of neurons, at about $0.3N$ patterns for the pseudoinverse rule and $0.5N$ for the optimized rule.
  • Overlapping memories work: every pattern activates half the neurons on average, so any pair shares about 50% of its active neurons, and the network still separates them.
  • Pattern completion is robust at low load (up to 30% of pattern bits flipped) and degrades gracefully to zero tolerance as the load approaches capacity.
  • The construction requires only inhibitory recurrent connections, so the mechanism is compatible with circuits that lack long-range excitatory recurrence.

Reading between the lines

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

  • A natural next test is whether the same hypercube recipe can host dense associative memories with exponential capacity; the paper does not explore higher-order vertices or beyond-pairwise interactions.
  • The equal-rate assumption predicts that firing-rate heterogeneity among active neurons shrinks capacity; injecting rate variance into $\eta_\mu$ is a direct stress test the paper leaves undone.
  • Because the optimal self-inhibition $\gamma$ depends on load, a biological implementation would need an activity-dependent reset or adaptation strength, an implicit constraint the authors do not discuss.
  • The encoding $N=2K$ costs two neurons per latent dimension; an asymmetric or non-orthogonal encoder could change the capacity coefficient, but the paper does not analyze that trade-off.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a geometric framework for storing overlapping binary memories in low-rank, all-inhibitory spiking networks. The authors map memory patterns to vertices of a hypercube in a latent space, design a rank constraint W = ED with an encoder E = [I_K; -I_K], and introduce Hebbian, pseudoinverse, and optimized decoders intended to satisfy the fixed-point condition D eta_mu = xi_mu. They report that the pseudoinverse and optimized rules store overlapping patterns, achieve an approximately linear storage capacity p/N around 0.3 and 0.5 in simulations, and show pattern completion, including an MNIST example. The paper also provides simulation code in a public repository.

Significance. If the central claims hold, the paper offers a scalable constructive recipe for spiking associative memory with overlapping patterns, directly connecting Hopfield-type memory to the latent-boundary framework of spiking networks. The geometric viewpoint is interpretable, the code is released, and the simulations demonstrate stable recall and pattern completion in networks up to N=1600. However, the analytical stability condition is only checked against an assumed equal-rate neural pattern rather than measured spiking rates, the capacity thresholds are empirical and come without error bars, and a normalization inconsistency with the parameter kappa undermines the stated derivation of the pseudoinverse rule. These issues bear directly on the paper's two headline claims.

major comments (3)
  1. [Sec. 2.2.3, Eqs. (10)-(11)] The constant kappa in eta_mu = ReLU(kappa E xi_mu) is never assigned a value. Combining the pseudoinverse decoder with this definition gives D_pinv eta_mu = kappa xi_mu (using E^T ReLU(E xi) = xi), which equals xi_mu only if kappa = 1 and the vertex coordinate c is normalized. As written, Eq. (8) is therefore not satisfied by the pseudoinverse rule for a general kappa, and the latent fixed point would be kappa xi_mu rather than xi_mu. The authors should state the normalization of kappa and c, or rescale the decoder accordingly.
  2. [Sec. 3.2, Fig. 4b] The headline claim of linear storage capacity is supported only by reading thresholds off a single overlap-versus-load curve at N=400, 800, and 1600, with no error bars, no distribution over pattern draws, and no analytical derivation of the p/N values of 0.3 and 0.5. The dimension-counting argument in Sec. 2.2.3 is only a necessary condition (p <= 2K = N) and does not establish why the spiking dynamics should remain stable up to the reported loads. The authors should either provide a fixed-point analysis of the designed decoder or add a systematic capacity measurement with confidence intervals and a defined success criterion.
  3. [Sec. 2.2.3, Fig. 3e, Appendix C.1] The verification of Eq. (8) in Fig. 3e uses the assumed pattern eta from Eq. (11), not the time-averaged firing rates of the simulated network. For the Hebbian and pseudoinverse rules, the self-connection gamma is imposed post hoc, so the implemented weight matrix is no longer W = ED and the exact constraint D eta = xi is broken in the simulated network. Since the spiking dynamics are oscillatory and gamma is tuned for each load (Fig. 4c), the claim that stable attractors sit exactly at the hypercube vertices is not directly established for the actual spiking system. The authors should measure the time-averaged firing rates at a recalled vertex and show that D<r> is approximately xi, or explain why the post-hoc self-connection does not affect the fixed-point condition.
minor comments (4)
  1. [Sec. 2.2.3, Eq. (12)] The sentence 'Given that D has KN = 2K^2 elements, this suggests that the linear system can only have a solution provided p <= 2K' is a heuristic necessary-condition argument, not a proof of feasibility for the specific ReLU-defined eta matrix; the text should label it as such.
  2. [Appendix C.2] The extra 0th latent dimension is introduced only in the appendix, while the main text describes a (K+1)-dimensional boundary before Eq. (7) defines a K-dimensional encoder; the relation between the rank-K latent dynamics and the rank-(K+1) weight matrix should be clarified in the main text.
  3. [Figure 3e] The caption does not define the quantity on the y-axis or explain the 'horizontal line indicates precise stability'; please state what is being plotted and how the horizontal line is determined.
  4. [Sec. 1] The introduction cites prior spiking associative memory models but does not quantify the memory loads they achieve; adding a brief quantitative comparison would help support the scalability claim.

Circularity Check

3 steps flagged · score 4.0 of 10

Analytic stability check reuses the fitted constraint; central capacity and recall claims rest on independent spiking simulations.

  1. self definitional [Section 2.2.3 (Eq. 12) and Section 3.1 (Fig. 3e)]
    "we formulate the linear system in Equation (8) as a least squares optimization problem in order to obtain a setting of the matrix D that stabilizes all memory patterns, which we write as Dopt = argmin_D ||D||2_2 subject to Dη = ξ, where we are assuming η as defined in Equation (11)."

    For the optimized rule, D is explicitly fitted to make Dη=ξ for the assumed pattern activity η=ReLU(κEξ). The later statement that the pseudoinverse and optimized rules 'more precisely satisfy the constraints' is therefore a report of the construction, not an independent check of stability. The same equation is used as both objective and verification. The non-circular evidence is the spiking simulation, which tests whether actual spike-filtered activity realizes the assumed η rather than simply checking the fitted identity.

  2. self citation load bearing [Section 2.2.4 and Appendix C.2]
    "Related work has shown that the boundary dynamics will be stable if the connectivity is all-inhibitory ... We refer the reader to Podlaski and Machens (2024), where such sign constraints are discussed in more detail."

    The stability of the all-inhibitory boundary is used to justify the extra latent dimension that makes the hypercube vertices stable, and this stability argument is imported from the authors' own prior work rather than proved here. The citation is not merely bibliographic; it supplies the sign-constraint theorem on which the manifold construction rests. However, the paper's own spiking simulations provide independent confirmation that the constructed networks are stable, so the self-citation is supportive rather than the only evidence.

1 more flagged steps
  1. fitted input called prediction [Section 3.2 (Fig. 4c) and Appendix C.1]
    "Empirically, we found that γ had to be optimized to promote stable recall ability, which was done numerically (see Figure 4c)."

    The self-connection strength γ is tuned separately for each network load to make recall work, and the capacity thresholds (p/N≈0.3 and 0.5) are then read from simulations using those tuned values. Because the free parameter is fit on the same recall task used to define capacity, the numerical capacity values are not parameter-free predictions. This is hyperparameter fitting rather than a circular derivation, and the linear scaling itself is not forced by the fit, so it contributes only mildly.

full rationale

The construction of D from Eq. (12) makes the optimized rule satisfy Dη=ξ by definition for the assumed η=ReLU(κEξ); Fig. 3e is therefore a consistency check of the fit rather than an independent test. Similarly, the all-inhibitory stability argument is taken from the authors' prior work (Podlaski and Machens, 2024). These are genuine circularity-adjacent steps, but they do not by themselves establish the paper's substantive claims. The reported recall, capacity scaling, and pattern completion are outcomes of spiking network simulations (Figs. 3a-b, 4a-d, 4e), which are external to the fitting procedure: the network either returns to the stored vertex or not, and the p/N thresholds are measured, not derived from the constraint. The capacity values also fall below the naive feasibility bound p≤2K, so they are not logically forced by the construction. The main unaddressed weakness is that the assumed equal-rate pattern η is never compared with measured time-averaged firing rates, and the kappa factor in Eq. (11) is not set to unity in the pseudoinverse rule as written, making the analytic fixed-point condition inconsistent; these are correctness risks rather than circular reductions. Overall, the central empirical claims have independent content, so the paper is only mildly circular.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central construction rests on a chain of design choices: the low-rank decomposition W = ED, the hypercube encoder E = [I; -I], the decoder fitted to D eta = xi, the equal-rate assumption for eta, the extra 0th latent dimension for inhibition, and the per-load tuning of gamma. Each is a choice made to make the construction work rather than a necessity derived from first principles. The main external input is the boundary-stability framework of Podlaski and Machens (2024), and the main internal heuristic is the dimension-counting argument for the feasibility of the optimized rule.

free parameters (3)
  • gamma (self-connection strength) = optimized numerically per load (Fig 4c)
    Appendix C.1: W_ii = -gamma for all i, and 'empirically, we found that gamma had to be optimized to promote stable recall ability.' The value is tuned for each network load and directly affects the attractor stability results.
  • kappa (target firing rate of active neurons) = positive constant, unspecified
    Eq (11) defines eta_mu = ReLU(kappa E xi_mu). kappa sets the scale of the assumed neural activity pattern and therefore the scale of the decoder solving Eq (8); it is treated as a reasonable assumption rather than derived.
  • c (hypercube vertex coordinate) = set via thresholds T and input Iext
    Section 2.2.1: c is the distance from the origin to each vertex; it depends on T and Iext(t), which are free construction choices that set the shape and position of the polyhedron.
assumptions (4)
  • domain assumption Spiking dynamics of rank-K all-inhibitory LIF networks are confined to a convex, piecewise-linear latent boundary and are stable when decoders point in subthreshold directions.
    Imported from Podlaski and Machens (2024); used in Section 2.1 and Appendix C.2 to justify the manifold picture and the all-inhibitory sign construction, not re-derived here.
  • standard math The latent dynamics reduce to y_dot = -y + D * I_<=0(Ey - T + Iext) (Eq 6), with spikes causing instantaneous jumps along decoder columns.
    Derived in Appendix A from LIF equations with delta-pulse synapses and exponential filtering; relies on the standard convex-analysis indicator function.
  • ad hoc to paper Each stored pattern's neural activity is eta_mu = ReLU(kappa E xi_mu), i.e., all active neurons fire at the same rate kappa.
    Eq (11); made 'to make this problem tractable'; not validated against the spiking network's actual active-neuron rates, and it is load-bearing for the decoder optimization.
  • domain assumption A solution to Eq (12) exists for p <= 2K by counting Kp equations against KN = 2K^2 unknowns.
    Section 2.2.3; the argument is a dimension count, not a proof for the specific correlated eta matrix produced by ReLU(kappa E xi), leaving feasibility to be checked empirically.
invented entities (1)
  • Extra 0th latent dimension
    purpose: Added to force all-inhibitory connectivity by setting E_i0 = 1 and D_0j = -alpha with alpha = max(W_ij), thereby stabilizing the boundary dynamics (Appendix C.2).
    An extra latent coordinate is introduced ad hoc to convert the learned weights into an all-inhibitory weight matrix; it is an architectural device with no external falsifiable handle beyond the paper's own simulations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Storing overlapping associative memories on latent manifolds in low-rank spiking networks." pith.science (2026). https://pith.science/paper/PDHCHBD2

@misc{pith2026241117485,
  author       = {Pith},
  title        = {Pith review of: Storing overlapping associative memories on latent manifolds in low-rank spiking networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PDHCHBD2}},
  note         = {Machine review of arXiv:2411.17485}
}
read the original abstract

Associative memory architectures such as the Hopfield network have long been important conceptual and theoretical models for neuroscience and artificial intelligence. However, translating these abstract models into spiking neural networks has been surprisingly difficult. Indeed, much previous work has been restricted to storing a small number of primarily non-overlapping memories in large networks, thereby limiting their scalability. Here, we revisit the associative memory problem in light of recent advances in understanding spike-based computation. Using a recently-established geometric framework, we show that the spiking activity for a large class of all-inhibitory networks is situated on a low-dimensional, convex, and piecewise-linear manifold, with dynamics that move along the manifold. We then map the associative memory problem onto these dynamics, and demonstrate how the vertices of a hypercubic manifold can be used to store stable, overlapping activity patterns with a direct correspondence to the original Hopfield model. We propose several learning rules, and demonstrate a linear scaling of the storage capacity with the number of neurons, as well as robust pattern completion abilities. Overall, this work serves as a case study to demonstrate the effectiveness of using a geometrical perspective to design dynamics on neural manifolds, with implications for neuroscience and machine learning.

Figures

Figures reproduced from arXiv: 2411.17485 by the authors.

Figure 1
Figure 1. The boundaries and dynamics of rank-constrained spiking networks. a: Each neu￾ron’s spike threshold forms a linear boundary in latent space that separates sub￾threshold (Vi < Ti) from suprathreshold (Vi > Ti) areas, with Ei being the ith row of the encoder matrix E. b: A population of neurons forms a convex, piecewise￾linear boundary between all neurons being subthreshold and suprathreshold areas. c: Dynamics are de… view at source ↗
Figure 2
Figure 2. Stabilizing vertices on a hypercubic boundary. a: A (K + 1) = 3-dimensional boundary composed of a hypercube (square), plus an additional open dimension that enables stable dynamics. b: 2d slice through panel (a) reveals a (K = 2)-d square boundary made of N = 2K = 4 neurons, and with 2K = 4 vertices as possible memory patterns. c-e: Close-up of the vertex at (c, −c), demonstrating a symmetric learning rule (c), asy… view at source ↗
Figure 3
Figure 3. Spiking associative memory in small networks. a: Example unsuccessful recall dynamics of the Hebbian network. b: Example recall dynamics of the optimized network, with successful recall (see [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Assessing storage capacity and pattern completion in larger networks. a: Demon￾stration of the attractor dynamics for a network of N = 100 neurons storing p = 40 patterns with the optimized learning rule. b,c: Average overlap across all patterns (b) and optimal self-co…
Figure 5
Figure 5. Figure 5: Demonstration of the small network simulation of N = 20 neurons for the pseu￾doinverse learning rule, showing successful pattern completion and recall (com￾pare with [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 41 canonical work pages

  1. [1]

    A learning algorithm for boltzmann machines

    David H Ackley, Geoffrey E Hinton, and Terrence J Sejnowski. A learning algorithm for boltzmann machines. Cognitive science, 9 0 (1): 0 147--169, 1985

    1985

  2. [2]

    Associative memory neural network with low temporal spiking rates

    Daniel J Amit and Alessandro Treves. Associative memory neural network with low temporal spiking rates. Proceedings of the National Academy of Sciences, 86 0 (20): 0 7871--7875, 1989

    work page 1989

  3. [3]

    Shaping dynamics with multiple populations in low-rank recurrent networks

    Manuel Beiran, Alexis Dubreuil, Adrian Valente, Francesca Mastrogiuseppe, and Srdjan Ostojic. Shaping dynamics with multiple populations in low-rank recurrent networks. Neural computation, 33 0 (6): 0 1572--1615, 2021

    work page 2021

  4. [4]

    Predictive coding of dynamical variables in balanced spiking networks

    Martin Boerlin, Christian K Machens, and Sophie Den \`e ve. Predictive coding of dynamical variables in balanced spiking networks. PLoS computational biology, 9 0 (11): 0 e1003258, 2013

    work page 2013

  5. [5]

    Convex optimization

    Stephen Boyd and Lieven Vandenberghe. Convex optimization. Cambridge university press, 2004

    2004

  6. [6]

    Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons

    Nicolas Brunel. Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons. Journal of computational neuroscience, 8: 0 183--208, 2000

    work page 2000

  7. [7]

    The geometry of robustness in spiking neural networks

    Nuno Calaim, Florian A Dehmelt, Pedro J Gon c alves, and Christian K Machens. The geometry of robustness in spiking neural networks. Elife, 11: 0 e73276, 2022

    work page 2022

  8. [8]

    Computational principles of memory

    Rishidev Chaudhuri and Ila Fiete. Computational principles of memory. Nature neuroscience, 19 0 (3): 0 394--403, 2016

    work page 2016

Show all 47 references
  1. [9]

    Neural population geometry: An approach for understanding biological and artificial neural networks

    SueYeon Chung and Larry F Abbott. Neural population geometry: An approach for understanding biological and artificial neural networks. Current opinion in neurobiology, 70: 0 137--144, 2021

    2021

  2. [10]

    Geometrical and statistical properties of systems of linear inequalities with applications in pattern recognition

    Thomas M Cover. Geometrical and statistical properties of systems of linear inequalities with applications in pattern recognition. IEEE transactions on electronic computers, 1 0 (3): 0 326--334, 1965

    1965

  3. [11]

    Benchmarks for progress in neuromorphic computing

    Mike Davies. Benchmarks for progress in neuromorphic computing. Nature Machine Intelligence, 1 0 (9): 0 386--388, 2019

    2019

  4. [12]

    The role of population structure in computations through neural dynamics

    Alexis Dubreuil, Adrian Valente, Manuel Beiran, Francesca Mastrogiuseppe, and Srdjan Ostojic. The role of population structure in computations through neural dynamics. Nature neuroscience, 25 0 (6): 0 783--794, 2022

    2022

  5. [13]

    A unified approach to building and controlling spiking attractor networks

    Chris Eliasmith. A unified approach to building and controlling spiking attractor networks. Neural computation, 17 0 (6): 0 1276--1314, 2005

    2005

  6. [14]

    Neural engineering: Computation, representation, and dynamics in neurobiological systems

    Chris Eliasmith and Charles H Anderson. Neural engineering: Computation, representation, and dynamics in neurobiological systems. MIT press, 2003

    2003

  7. [15]

    Robust computation with rhythmic spike patterns

    E Paxon Frady and Friedrich T Sommer. Robust computation with rhythmic spike patterns. Proceedings of the National Academy of Sciences, 116 0 (36): 0 18050--18059, 2019

    2019

  8. [16]

    Neural manifolds for the control of movement

    Juan A Gallego, Matthew G Perich, Lee E Miller, and Sara A Solla. Neural manifolds for the control of movement. Neuron, 94 0 (5): 0 978--984, 2017

    2017

  9. [17]

    The space of interactions in neural network models

    Elizabeth Gardner. The space of interactions in neural network models. Journal of physics A: Mathematical and general, 21 0 (1): 0 257, 1988

    1988

  10. [18]

    Associative memory in a network of ‘spiking’ neurons

    Wulfram Gerstner and J Leo van Hemmen. Associative memory in a network of ‘spiking’ neurons. Network: Computation in Neural Systems, 3 0 (2): 0 139--164, 1992

    1992

  11. [19]

    Neuronal dynamics: From single neurons to networks and models of cognition

    Wulfram Gerstner, Werner M Kistler, Richard Naud, and Liam Paninski. Neuronal dynamics: From single neurons to networks and models of cognition. Cambridge University Press, 2014

    2014

  12. [20]

    Population adaptation in efficient balanced networks

    Gabrielle J Gutierrez and Sophie Den \`e ve. Population adaptation in efficient balanced networks. ELife, 8: 0 e46926, 2019

    2019

  13. [21]

    Introduction to the theory of neural computation, 1991

    John Hertz, Anders Krogh, and Richard G Palmer. Introduction to the theory of neural computation, 1991

    1991

  14. [22]

    Neural networks and physical systems with emergent collective computational abilities

    John J Hopfield. Neural networks and physical systems with emergent collective computational abilities. Proceedings of the national academy of sciences, 79 0 (8): 0 2554--2558, 1982

    1982

  15. [23]

    Interpreting neural computations by examining intrinsic and embedding dimensionality of neural activity

    Mehrdad Jazayeri and Srdjan Ostojic. Interpreting neural computations by examining intrinsic and embedding dimensionality of neural activity. Current opinion in neurobiology, 70: 0 113--120, 2021

    2021

  16. [24]

    Finding the engram

    Sheena A Josselyn, Stefan K \"o hler, and Paul W Frankland. Finding the engram. Nature Reviews Neuroscience, 16 0 (9): 0 521--534, 2015

    2015

  17. [25]

    Associative recall of memory without errors

    I Kanter and Haim Sompolinsky. Associative recall of memory without errors. Physical Review A, 35 0 (1): 0 380, 1987

    1987

  18. [26]

    A new frontier for hopfield networks

    Dmitry Krotov. A new frontier for hopfield networks. Nature Reviews Physics, 5 0 (7): 0 366--367, 2023

    2023

  19. [27]

    Dense associative memory for pattern recognition

    Dmitry Krotov and John J Hopfield. Dense associative memory for pattern recognition. Advances in neural information processing systems, 29, 2016

    2016

  20. [28]

    In search of the engram

    Karl S Lashley. In search of the engram. Society for Experimental Biology, Physiological mechanisms in animal behavior, pages 454--482, 1950

    1950

  21. [29]

    Gradient-based learning applied to document recognition

    Yann LeCun, L \'e on Bottou, Yoshua Bengio, and Patrick Haffner. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86 0 (11): 0 2278--2324, 1998

    1998

  22. [30]

    Formation and maintenance of neuronal assemblies through synaptic plasticity

    Ashok Litwin-Kumar and Brent Doiron. Formation and maintenance of neuronal assemblies through synaptic plasticity. Nature communications, 5 0 (1): 0 5319, 2014

    2014

  23. [31]

    Networks of spiking neurons can emulate arbitrary hopfield nets in temporal coding

    Wolfgang Maass and Thomas Natschl \"a ger. Networks of spiking neurons can emulate arbitrary hopfield nets in temporal coding. Network: Computation in Neural Systems, 8 0 (4): 0 355, 1997

    1997

  24. [32]

    Understanding spiking networks through convex optimization

    Allan Mancoo, Sander Keemink, and Christian K Machens. Understanding spiking networks through convex optimization. Advances in Neural Information Processing Systems, 33: 0 8824--8835, 2020

    2020

  25. [33]

    Three types of remapping with linear decoders: a population-geometric perspective

    Guillermo Mart \' n-S \'a nchez, Christian K Machens, and William F Podlaski. Three types of remapping with linear decoders: a population-geometric perspective. bioRxiv, pages 2025--03, 2025

    2025

  26. [34]

    Linking connectivity, dynamics, and computations in low-rank recurrent neural networks

    Francesca Mastrogiuseppe and Srdjan Ostojic. Linking connectivity, dynamics, and computations in low-rank recurrent neural networks. Neuron, 99 0 (3): 0 609--623, 2018

    2018

  27. [35]

    Content-addressable memory with spiking neurons

    R Mueller and Andreas VM Herz. Content-addressable memory with spiking neurons. Physical Review E, 59 0 (3): 0 3330, 1999

    1999

  28. [36]

    Information storage and retrieval in spin-glass like neural networks

    L Personnaz, I Guyon, and G Dreyfus. Information storage and retrieval in spin-glass like neural networks. Journal de Physique Lettres, 46 0 (8): 0 359--365, 1985

    1985

  29. [37]

    Approximating nonlinear functions with latent boundaries in low-rank excitatory-inhibitory spiking networks

    William F Podlaski and Christian K Machens. Approximating nonlinear functions with latent boundaries in low-rank excitatory-inhibitory spiking networks. Neural Computation, 36 0 (5): 0 803--857, 2024

    2024

  30. [38]

    High capacity and dynamic accessibility in associative memory networks with context-dependent neuronal and synaptic gating

    William F Podlaski, Everton J Agnes, and Tim P Vogels. High capacity and dynamic accessibility in associative memory networks with context-dependent neuronal and synaptic gating. Physical Review X, 15 0 (1): 0 011057, 2025

    2025

  31. [39]

    Engineering recurrent neural networks from task-relevant manifolds and dynamics

    Eli Pollock and Mehrdad Jazayeri. Engineering recurrent neural networks from task-relevant manifolds and dynamics. PLoS computational biology, 16 0 (8): 0 e1008128, 2020

    2020

  32. [40]

    Hopfield networks is all you need

    Hubert Ramsauer, Bernhard Sch \"a fl, Johannes Lehner, Philipp Seidl, Michael Widrich, Lukas Gruber, Markus Holzleitner, Thomas Adler, David Kreil, Michael K Kopp, et al. Hopfield networks is all you need. In International Conference on Learning Representations, 2020

    2020

  33. [41]

    Content addressable memory without catastrophic forgetting by heteroassociation with a fixed scaffold

    Sugandha Sharma, Sarthak Chandra, and Ila Fiete. Content addressable memory without catastrophic forgetting by heteroassociation with a fixed scaffold. In International Conference on Machine Learning, pages 19658--19682. PMLR, 2022

    2022

  34. [42]

    Associative memory in networks of spiking neurons

    Friedrich T Sommer and Thomas Wennekers. Associative memory in networks of spiking neurons. Neural networks, 14 0 (6-7): 0 825--834, 2001

    2001

  35. [43]

    Computational analysis of the role of the hippocampus in memory

    Alessandro Treves and Edmund T Rolls. Computational analysis of the role of the hippocampus in memory. Hippocampus, 4 0 (3): 0 374--391, 1994

    1994

  36. [44]

    Computation through neural population dynamics

    Saurabh Vyas, Matthew D Golub, David Sussillo, and Krishna V Shenoy. Computation through neural population dynamics. Annual review of neuroscience, 43 0 (1): 0 249--275, 2020

    2020

  37. [45]

    The remarkable robustness of surrogate gradient learning for instilling complex function in spiking neural networks

    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

    2021

  38. [46]

    Diverse synaptic plasticity mechanisms orchestrated to form and retrieve memories in spiking neural networks

    Friedemann Zenke, Everton J Agnes, and Wulfram Gerstner. Diverse synaptic plasticity mechanisms orchestrated to form and retrieve memories in spiking neural networks. Nature communications, 6 0 (1): 0 6922, 2015

    2015

  39. [47]

    Visualizing a joint future of neuroscience and neuromorphic engineering

    Friedemann Zenke, Sander M Boht \'e , Claudia Clopath, Iulia M Com s a, Julian G \"o ltz, Wolfgang Maass, Timoth \'e e Masquelier, Richard Naud, Emre O Neftci, Mihai A Petrovici, et al. Visualizing a joint future of neuroscience and neuromorphic engineering. Neuron, 109 0 (4):...

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.