Pith. sign in

REVIEW 3 major objections 6 minor 69 references

Hopfield Networks as Models of Emergent Function in Biology

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read This review argues that Hopfield networks—originally a model of associative memory in neural circuits—are a unifying lens for emergent biological function across cell fate, self-assembly, and neural representation.

desk verdict A potentially useful review of Hopfield models in biology, but the math tutorial has load-bearing equation errors (wrong correlation matrix, missing 1/N, and a wrong β→0 limit) that need fixing first. read the letter →

arxiv 2506.13076 v1 pith:ZHSXPQKF submitted 2025-06-16 physics.bio-ph cond-mat.dis-nncond-mat.softcond-mat.stat-mechq-bio.NC

classification physics.bio-phcond-mat.dis-nncond-mat.softcond-mat.stat-mechq-bio.NC
keywords Hopfieldnetworksassociativememoryemergentbiologicalfunctioncellulardifferentiationepigeneticself-assemblyneuralrepresentationsenergylandscape
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

This review argues that Hopfield networks—originally a model of associative memory in neural circuits—capture a common mechanism behind emergent biological function: high-dimensional variables whose dynamics flow to a small set of stored patterns can retrieve a whole 'memory' from a partial or noisy cue. The authors present the mathematics of the classic Hopfield network, the projection method for correlated patterns, and exponential modern Hopfield networks, then translate the dynamics into three complementary pictures: separating signal from noise, projecting onto the subspace spanned by stored patterns, and descending a parabolic energy landscape in order-parameter space. They apply these pictures to stable cell fates and epigenetic memory, multicomponent molecular self-assembly, and hippocampal spatial representations. The payoff, if the lens is right, is a single set of formal objects—stored patterns, overlaps, attractors, landscape gradients—that organizes phenomena across scales and gives a practical coordinate system for high-dimensional biological data.

What carries the argument

The load-bearing object is the order-parameter vector $m^\mu = \frac{1}{N}\sum_i \xi_i^\mu x_i$, with the decorrelated generalization $\bar{m}^\mu = \sum_\nu g^{\mu\nu} m_\nu$ for correlated patterns, where $g_{\mu\nu}$ is the pattern overlap matrix. This vector converts the $N$-dimensional neuron state into $P$-dimensional pattern-space coordinates, and every result the review needs—the energy $E = -\frac{N}{2}\sum_\mu \bar{m}^\mu m_\mu$, the update rule, storage capacity, and the projection interpretation—is stated in this space. The projection matrix $\mathbf{P}_{ij} = \sum_{\mu,\nu} \xi_i^\mu g^{\mu\nu} \xi_j^\nu$ is the same object as the projection-rule couplings, making the 'dynamics as projection' picture literal. The second key mechanism is the nonlinearity (sign in the classic model, softmax in the modern one) that rectifies gradient descent on the parabolic landscape and, when steep, produces the exponential storage capacity of modern Hopfield networks.

What would settle it

Follow Eq. 6 with two orthogonal patterns, say $\xi^1=(1,-1)$ and $\xi^2=(1,1)$; the proposed $g_{\mu\nu}$ is not the identity matrix, so the projection identity $\mathbf{P}^2=\mathbf{P}$ fails, whereas the Gram matrix version gives the identity. Separately, simulate the exponential update at small $\beta$ (e.g., $\beta=0.01$): the softmax is nearly uniform and a stored pattern is not a fixed point, contradicting the paper's fixed-point claim; at large $\beta$ (e.g., $\beta=1000$) the stored patterns are fixed points, confirming that $\beta \to \infty$, not $\beta \to 0$, is the retrieval limit.

Watch

Extended reading notes

Core claim

The central claim is that Hopfield models offer 'a striking example of how simple dynamical rules in high-dimensional systems can give rise to robust, emergent function in biology.' The review's contribution is to make that claim teachable and transferable: it derives the classic update rule and Lyapunov energy, introduces the generalized order parameters $\bar{m}^\mu = \sum_\nu g^{\mu\nu} m_\nu$ for correlated patterns, and shows that the energy in these coordinates is an inverted $P$-dimensional parabola $E = -\frac{N}{2}\sum_\mu \bar{m}^\mu m_\mu$. Three interpretations—noise discrimination, geometric projection, landscape descent—are presented as complementary readings of the same dynamics, and each biological application (cell differentiation, self-assembly, spatial cognition) is mapped onto the same elements: state variables, couplings, stored patterns, and retrieval dynamics.

Load-bearing premise

The load-bearing premise is that the tutorial's derivations are correct as written, and two spots currently mislead: Eq. 6 writes the pattern correlation as a double sum over coordinates instead of the standard Gram matrix $(1/N)\sum_i \xi_i^\mu \xi_i^\nu$, and Eq. 13 calls $\beta \to 0$ the zero-temperature limit when $\beta \to 0$ is actually the high-temperature limit of the softmax (the needed limit is $\beta \to \infty$).

Editorial extensions

If this is right

  • If the lens is correct, the same mathematical objects—stored patterns, overlaps, energy landscapes—organize gene regulation, molecular assembly, and spatial memory, so insights from one field transfer to the others.
  • Hopfield-style order parameters give a parameter-free, interpretable coordinate system for single-cell gene-expression atlases, an alternative to stochastic embeddings for classifying cells and tracking fate transitions.
  • The contrast between linear capacity in classic Hopfield networks and exponential capacity in softmax-based modern networks explains why stronger nonlinearities allow vastly more stored states, a fact relevant to how biological systems might store many cell types or structures.
  • The projection method shows that correlated stored patterns can be retrieved if couplings correct for their overlap, meaning interference from correlated cell types or molecular structures is a solvable engineering problem rather than a fundamental limit.
  • Because classic and modern Hopfield networks are both limits of a common bipartite visible-hidden Lagrangian dynamics, results about one formulation can be translated into the other, including the connection to restricted Boltzmann machines and attention.

Reading between the lines

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

  • The review leaves implicit a stronger program: if order parameters place every cell on a fate simplex, then differentiation and reprogramming become paths along simplex coordinates, which could be tested by measuring whether scRNA-seq trajectories stay on that simplex across perturbations.
  • The self-assembly analogy suggests a quantitative capacity rule the review does not derive: the number of reliably assemblable structures should be controlled by the effective overlap between target structures, so measuring that overlap from component interaction data would predict when a molecular library fails.
  • The link between modern Hopfield networks, softmax attention, and hippocampal maps invites a cross-domain prediction: replay-like retrieval in biological memory and attention-like readout in machine learning may be the same attractor mechanism, which could be tested by comparing model dynamics to place-cell reactivation data.
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 / 6 minor

Summary. This review paper argues that Hopfield networks provide a unifying conceptual lens for emergent function in biology. It gives a pedagogical introduction to classical Hopfield networks, the projection method for correlated patterns, and exponential modern Hopfield networks, and then presents three complementary interpretations of the dynamics: signal retrieval, projection onto a pattern subspace, and descent on an energy landscape. The second half surveys applications to cell-fate decisions and epigenetics, molecular self-assembly, and hippocampal spatial representations, and concludes that Hopfield models illustrate how simple high-dimensional dynamical rules can yield robust biological function.

Significance. If the mathematical presentation is corrected, this review would be a valuable pedagogical entry point for biophysicists. The paper's strength is its synthesis: it connects classical and modern Hopfield models through the common language of order parameters, offers three physically motivated interpretations, and illustrates each with figures and concrete biological examples. The survey of applications is timely and covers an appropriate range of systems. The self-citations to the authors' prior order-parameter work are used as application examples rather than as circular support, and the review makes no overreaching claim to new results or fitted parameters. The main risk is that the tutorial sections contain several errors in core equations; because the paper's stated aim is a self-contained introduction, these errors are load-bearing and must be fixed before the review can serve its intended purpose.

major comments (3)
  1. [Sec. 2.3, Eq. (6)] The correlation matrix is defined as g_{\mu\nu} = \sum_{ij} \xi^\mu_i \xi^\nu_j. This double sum factorizes as (\sum_i \xi^\mu_i)(\sum_j \xi^\nu_j), a rank-one matrix determined by the pattern means, not the pattern overlaps. For unbiased random patterns it is approximately zero, so the inverse in Eq. (7) is ill-defined. The standard Gram matrix is g_{\mu\nu} = (1/N)\sum_i \xi^\mu_i \xi^\nu_i (or the unnormalized version \sum_i). Because Eqs. (8), (9), (13), (14), (16), and (18) all depend on g, this single definition propagates through the projection-method and modern-Hopfield derivations and must be corrected first.
  2. [Sec. 2.4, Eq. (13) and following sentence] The fixed-point argument states that 'in the zero temperature limit \beta\to 0' the one-hot order-parameter vector is a fixed point. Since \beta is the inverse temperature in the softmax, Eq. (12), the zero-temperature limit is \beta\to\infty, not \beta\to 0. In fact, \beta\to 0 gives \sigma_\mu(\beta m_\mu)\to 1/P for all \mu, so the one-hot vector is not a fixed point in that limit. The sharpening to a one-hot distribution occurs for \beta\to\infty, provided the target pattern has the largest m_\mu. This is not a notational slip; it reverses the temperature dependence and directly undermines the subsequent claim that modern Hopfield networks retrieve the stored pattern they are closest to. Additionally, the derivation of Eq. (13) drops the factor N that comes from the definition m^\mu = (1/N)\sum_j \xi^\mu_j x_j in Eq. (4) and from the use of g in the contraction; this inconsistency should be resolved after Eq. (6) is corrected.
  3. [Sec. 3.2, Eq. (14)] The projection matrix is written as P_{ij} = \sum_{\mu,\nu} \xi^\mu_i g^{\mu\nu} \xi^\nu_j, but the coupling matrix in Eq. (8) includes an explicit 1/N factor: J_{ij} = (1/N)\sum_{\mu,\nu} \xi^\mu_i g^{\mu\nu} \xi^\nu_j. With the standard normalization of g, the projector onto the pattern subspace is P = (1/N)\Xi g^{-1} \Xi^T, and the 1/N is required for P^2 = P. As written, Eq. (14) does not satisfy P^2 = P, and the statement immediately below that 'the couplings J_{ij} defined in Eq. 8 are exactly the projection matrix J_{ij} = P_{ij}' is inconsistent with the printed equations. This affects the interpretation of the dynamics as projection in Eqs. (15) and (16).
minor comments (6)
  1. [Sec. 2.3, notation] The notation g_{\mu\nu} for the matrix and g^{\mu\nu} for its inverse is introduced in Eqs. (6) and (7), but the paper never explains the index-raising convention or why the same symbol appears with lower and upper indices in Eq. (9). A brief notational remark would help readers follow the projection-method derivation.
  2. [Sec. 2.4, after Eq. (13)] In the sentence describing the fixed point, '\sigma_\mu(\beta m_\mu)\approx 1 if \nu=\gamma' should read '\mu=\gamma' rather than '\nu=\gamma'.
  3. [Sec. 2.2, spurious states] The line defining odd mixtures of patterns reads '\pm \vec{\xi}^{\mu_1} \pm \vec{\xi}^{\mu_2} \pm \vec{\xi}^{\mu_2}', repeating pattern \mu_2; it should presumably involve three distinct patterns (or two if only two patterns are intended).
  4. [Sec. 2.3, Eq. (10)] Equation (10) appears to print two competing equalities in sequence, one containing g^{\mu\nu} and one without. Please check the intended expression; if the decorrelated order parameters are defined by Eq. (9), the energy should be written as - (N/2)\sum_{\mu} m^\mu m_\mu, with the index convention stated explicitly.
  5. [Abstract] The abstract contains a grammatical fragment: 'Hopfield networks a variety of biological setting' should be 'Hopfield networks in a variety of biological settings.'
  6. [Sec. 3.3, Eq. (18)] The text says the energy is 'an inverted P-dimensional parabola centered at \vec{m}=0'; this is true only if the metric g^{\mu\nu} is positive definite and the order-parameter space is equipped with the appropriate inner product. A one-sentence clarification would avoid confusion for readers encountering the projection method for the first time.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: standard pedagogical review; self-citations are illustrative, not load-bearing.

full rationale

This is a pedagogical review rather than an original derivation with fitted parameters or empirical predictions. The mathematical core (Hebbian couplings in Eq. 2, the projection rule in Eq. 8, and the exponential modern Hopfield update/energy in Eq. 11) is presented as a self-contained exposition of standard results (Hopfield 1982; Personnaz et al. 1985; Kanter & Sompolinsky 1987; Demircigil et al. 2017), so there is no fitted-input-called-prediction or definitional reduction. The only self-citations (Refs. 67 and 68, plus the adapted figure in Fig. 1C) are used as illustrative applications and to support the statement that order-parameter coordinates are useful on real scRNA-seq atlases; this rests on external data analysis and is not the load-bearing premise of the review, so it is independent support rather than circularity. The one notable technical defect is Eq. 13's 'zero temperature limit β→0,' which is inconsistent with β as inverse temperature (the softmax sharpens as β→∞); this is a correctness issue in the exposition, not a circular step. No derivation in the paper reduces by construction to its own input or to a self-citation chain.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The review introduces no free parameters and no new entities. The only hand-set values are simulation parameters in figures (T=0.25, beta=100 log P), which do not affect the analytical claims. The biological mappings in Section 4 are the least empirically grounded assumptions; the review presents them as established applications rather than testing them.

assumptions (6)
  • standard math Central limit theorem applied to the interference term in the classic Hopfield update.
    Used in Section 3.1 to show retrieval errors scale as sqrt(P/N). This is standard and does not depend on the biology.
  • domain assumption Stored patterns in the classic network are independent random +/-1 variables.
    Used in Sections 2.2 and 3.1 for the signal-noise calculation and the linear storage capacity result.
  • domain assumption The pattern correlation matrix in the projection method is invertible.
    Invoked in Eq. 7 and Eq. 8 in Section 2.3; the projection rule requires the inverse of g. The paper's Eq. 6 as written does not define a generally invertible matrix.
  • domain assumption Stable cell fates correspond to attractors of gene regulatory dynamics and can be represented as stored Hopfield patterns.
    Assumed in Section 4.1 when mapping gene expression to x and cell types to xi; the review cites modeling papers but does not validate this mapping with primary data.
  • domain assumption Molecular self-assembly can be described by pairwise interactions and target structures as stored patterns.
    Assumed in Section 4.2 following Murugan et al. and related work; the review notes nonspecific binding causes interference analogous to pattern noise.
  • domain assumption Hippocampal cognitive maps can be modeled as stored patterns retrieved by Hopfield-like dynamics.
    Assumed in Section 4.3; the review explicitly says a detailed account is beyond its scope.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hopfield Networks as Models of Emergent Function in Biology." pith.science (2026). https://pith.science/paper/ZHSXPQKF

@misc{pith2026250613076,
  author       = {Pith},
  title        = {Pith review of: Hopfield Networks as Models of Emergent Function in Biology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZHSXPQKF}},
  note         = {Machine review of arXiv:2506.13076}
}
read the original abstract

Hopfield models, originally developed to study memory retrieval in neural networks, have become versatile tools for modeling diverse biological systems in which function emerges from collective dynamics. In this review, we provide a pedagogical introduction to both classical and modern Hopfield networks from a biophysical perspective. After presenting the underlying mathematics, we build physical intuition through three complementary interpretations of Hopfield dynamics: as noise discrimination, as a geometric construction defining a natural coordinate system in pattern space, and as gradient-like descent on an energy landscape. We then survey recent applications of Hopfield networks a variety of biological setting including cellular differentiation and epigenetic memory, molecular self-assembly, and spatial neural representations.

Figures

Figures reproduced from arXiv: 2506.13076 by the authors.

Figure 2
Figure 2. The Hopfield model can perform accurate pattern retrieval below the [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

69 extracted references · 55 canonical work pages

  1. [1]

    Williams

    Herv´ e Abdi and Lynne J. Williams. Principal component analysis.WIREs Computational Statistics, 2(4):433–459, 2010. . URLhttps://wires.onlinelibrary.wiley.com/doi/abs/10. 1002/wics.101

  2. [2]

    Tolerance versus synaptic noise in dense associative memories

    Elena Agliari and Giordano De Marzo. Tolerance versus synaptic noise in dense associative memories. 2020. . URLhttps://arxiv.org/abs/2007.02849

  3. [3]

    Dense hebbian neural networks: A replica symmetric picture of supervised learning.Physica A: Statistical Mechanics and its Applications, 626:129076, September 2023

    Elena Agliari, Linda Albanese, Francesco Alemanno, Andrea Alessandrelli, Adriano Barra, Fosca Giannotti, Daniele Lotito, and Dino Pedreschi. Dense hebbian neural networks: A replica symmetric picture of supervised learning.Physica A: Statistical Mechanics and its Applications, 626:129076, September 2023. ISSN 0378-4371. . URLhttp://dx.doi.org/10.1016/j.ph...

  4. [4]

    The cell as a collection of protein machines: preparing the next generation of molecular biologists.cell, 92(3):291–294, 1998

    Bruce Alberts. The cell as a collection of protein machines: preparing the next generation of molecular biologists.cell, 92(3):291–294, 1998

  5. [5]

    Cambridge university press, 1989

    Daniel J Amit and Daniel J Amit.Modeling brain function: The world of attractor neural networks. Cambridge university press, 1989

  6. [7]

    Storing infinite numbers of patterns in a spin-glass model of neural networks.Physical Review Letters, 55(14):1530, 1985

    Daniel J Amit, Hanoch Gutfreund, and Haim Sompolinsky. Storing infinite numbers of patterns in a spin-glass model of neural networks.Physical Review Letters, 55(14):1530, 1985

  7. [8]

    A new mechanical approach to handle generalized hopfield neural networks.Neural Networks, 106:205–222, October 2018

    Adriano Barra, Matteo Beccaria, and Alberto Fachechi. A new mechanical approach to handle generalized hopfield neural networks.Neural Networks, 106:205–222, October 2018. ISSN 0893-6080. . URLhttp://dx.doi.org/10.1016/j.neunet.2018.07.010

  8. [9]

    Princeton University Press, 2012

    William Bialek.Biophysics: searching for principles. Princeton University Press, 2012

Show all 69 references
  1. [10]

    Waddington landscape for prototype learning in generalized hopfield networks

    Nacer Eddine Boukacem, Allen Leary, Robin Th´ eriault, Felix Gottlieb, Madhav Mani, and Paul Fran¸ cois. Waddington landscape for prototype learning in generalized hopfield networks. Physical Review Research, 6(3):033098, 2024

  2. [11]

    Liquid hopfield model: Retrieval and localization in multicomponent liquid mixtures.Proceedings of the National Academy of Sciences, 121(48):e2320504121, 2024

    Rodrigo Braz Teixeira, Giorgio Carugno, Izaak Neri, and Pablo Sartori. Liquid hopfield model: Retrieval and localization in multicomponent liquid mixtures.Proceedings of the National Academy of Sciences, 121(48):e2320504121, 2024

  3. [12]

    Learn- ing and inference in a lattice model of multicomponent condensates

    Cameron Chalk, Salvador Buse, Krishna Shrinivas, Arvind Murugan, and Erik Winfree. Learn- ing and inference in a lattice model of multicomponent condensates. In30th International Conference on DNA Computing and Molecular Programming (DNA 30)(2024), pages 5–1. Schloss Dagstuhl–...

  4. [13]

    The specious art of single-cell genomics.PLOS Computational Biology, 19(8):e1011288, 2023

    Tara Chari and Lior Pachter. The specious art of single-cell genomics.PLOS Computational Biology, 19(8):e1011288, 2023

  5. [14]

    On a model of associative memory with huge storage capacity.Journal of Statistical Physics, 168:288–299, 2017

    Mete Demircigil, Judith Heusel, Matthias L¨ owe, Sven Upgang, and Franck Vermet. On a model of associative memory with huge storage capacity.Journal of Statistical Physics, 168:288–299, 2017

  6. [15]

    Pattern recog- nition in the nucleation kinetics of non-equilibrium self-assembly.Nature, 625(7995):500–507, 2024

    Constantine Glen Evans, Jackson O’Brien, Erik Winfree, and Arvind Murugan. Pattern recog- nition in the nucleation kinetics of non-equilibrium self-assembly.Nature, 625(7995):500–507, 2024

  7. [16]

    Not just a colourful metaphor: modelling the landscape of cellular development using hopfield networks

    Atefeh Taherian Fard, Sriganesh Srihari, Jessica C Mar, and Mark A Ragan. Not just a colourful metaphor: modelling the landscape of cellular development using hopfield networks. NPJ systems biology and applications, 2(1):1–9, 2016

  8. [17]

    Bistability, bifurcations, and waddington’s epigenetic landscape.Current biology, 22(11):R458–R466, 2012

    James E Ferrell. Bistability, bifurcations, and waddington’s epigenetic landscape.Current biology, 22(11):R458–R466, 2012

  9. [18]

    Multiconnected neural network models.Journal of Physics A: Mathematical and General, 20(11):3453–3464, August 1987

    E Gardner. Multiconnected neural network models.Journal of Physics A: Mathematical and General, 20(11):3453–3464, August 1987. ISSN 1361-6447. . URLhttp://dx.doi.org/10. 1088/0305-4470/20/11/046

  10. [19]

    Jing Guo and Jie Zheng. Hopland: single-cell pseudotime recovery using continuous hopfield www.annualreviews.org • Hopfield Models 19 network-based modeling of waddington’s epigenetic landscape.Bioinformatics, 33(14):i102– i109, 2017

  11. [20]

    What is life? the physical aspect of the living cell: By erwin schr¨ odinger

    Alexander Haddow. What is life? the physical aspect of the living cell: By erwin schr¨ odinger. cambridge: University press. 1944, viii+ 91 pages, 1945

  12. [21]

    Durable alveolar engraftment of psc-derived lung epithelial cells into immunocompetent mice.Cell Stem Cell, 30(9):1217–1234, 2023

    Michael J Herriges, Maria Yampolskaya, Bibek R Thapa, Jonathan Lindstrom-Vautrin, Feiya Wang, Jessie Huang, Cheng-Lun Na, Liang Ma, McKenna M Montminy, Pushpinder Bawa, et al. Durable alveolar engraftment of psc-derived lung epithelial cells into immunocompetent mice.Cell Stem...

  13. [22]

    Neurons with graded response have collective computational properties like those of two-state neurons.Proceedings of the National Academy of Sciences, 81(10):3088–3092, 1984

    J J Hopfield. Neurons with graded response have collective computational properties like those of two-state neurons.Proceedings of the National Academy of Sciences, 81(10):3088–3092, 1984. . URLhttps://www.pnas.org/doi/abs/10.1073/pnas.81.10.3088

  14. [23]

    J. J. Hopfield, D. I. Feinstein, and R. G. Palmer. ‘unlearning’ has a stabilizing effect in collective memories.Nature, 304(5922):158–159, July 1983. ISSN 1476-4687. . URLhttp://dx.doi.org/ 10.1038/304158a0

  15. [24]

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

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

  16. [25]

    Neurodynamics of mental exploration.Proceedings of the National Academy of Sciences, 107(4):1648–1653, 2010

    John J Hopfield. Neurodynamics of mental exploration.Proceedings of the National Academy of Sciences, 107(4):1648–1653, 2010

  17. [26]

    Whatever happened to solid state physics?Annu

    John J Hopfield. Whatever happened to solid state physics?Annu. Rev. Condens. Matter Phys., 5(1):1–13, 2014

  18. [27]

    Hotelling

    H. Hotelling. Analysis of a complex of statistical variables into principal components.Journal of Educational Psychology, 24(6):417–441, September 1933. ISSN 0022-0663. . URLhttp: //dx.doi.org/10.1037/h0071325

  19. [28]

    The molecular and mathematical basis of waddington’s epigenetic landscape: A framework for post-darwinian biology?Bioessays, 34(2):149–157, 2012

    Sui Huang. The molecular and mathematical basis of waddington’s epigenetic landscape: A framework for post-darwinian biology?Bioessays, 34(2):149–157, 2012

  20. [29]

    The in vivo genetic program of murine primordial lung epithelial progenitors.Nature communications, 11(1):635, 2020

    Laertis Ikonomou, Michael J Herriges, Sara L Lewandowski, Robert Marsland III, Carlos Villacorta-Martin, Ignacio S Caballero, David B Frank, Reeti M Sanghrajka, Keri Dame, Maciej M Ka´ ndu la, et al. The in vivo genetic program of murine primordial lung epithelial progenitors....

  21. [30]

    Random features hopfield networks generalize retrieval to previously unseen examples.arXiv preprint arXiv:2407.05658, 2024

    Silvio Kalaj, Clarissa Lauditi, Gabriele Perugini, Carlo Lucibello, Enrico M Malatesta, and Matteo Negri. Random features hopfield networks generalize retrieval to previously unseen examples.arXiv preprint arXiv:2407.05658, 2024

  22. [31]

    Kanter and H

    I. Kanter and H. Sompolinsky. Associative recall of memory without errors.Physical Review A, 35(1):380–392, January 1987. ISSN 0556-2791. . URLhttp://dx.doi.org/10.1103/PhysRevA. 35.380

  23. [32]

    Enhancernet: a predictive model of cell identity dynamics through enhancer selection.Development, 151(19):dev202997, 2024

    Omer Karin. Enhancernet: a predictive model of cell identity dynamics through enhancer selection.Development, 151(19):dev202997, 2024

  24. [33]

    A new frontier for hopfield networks.Nature Reviews Physics, 5(7):366–367, 2023

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

  25. [34]

    Large associative memory problem in neurobiology and machine learning.arXiv preprint arXiv:2008.06996, 2020

    Dmitry Krotov and John Hopfield. Large associative memory problem in neurobiology and machine learning.arXiv preprint arXiv:2008.06996, 2020

  26. [35]

    Dense associative memory for pattern recognition.Ad- vances in neural information processing systems, 29, 2016

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

  27. [36]

    Epigenetic landscapes explain partially reprogrammed cells and identify key reprogramming genes.PLoS computational biology, 10(8): e1003734, 2014

    Alex H Lang, Hu Li, James J Collins, and Pankaj Mehta. Epigenetic landscapes explain partially reprogrammed cells and identify key reprogramming genes.PLoS computational biology, 10(8): e1003734, 2014

  28. [37]

    The exponential capacity of dense associative memories

    Carlo Lucibello and Marc M´ ezard. The exponential capacity of dense associative memories

  29. [38]

    Characterizing cancer subtypes as attractors of hopfield networks.Bioinformatics, 30(9):1273–1279, 2014

    Stefan R Maetschke and Mark A Ragan. Characterizing cancer subtypes as attractors of hopfield networks.Bioinformatics, 30(9):1273–1279, 2014

  30. [39]

    Umap: Uniform manifold approximation and 20 Yampolskaya and Mehta projection for dimension reduction.arXiv preprint arXiv:1802.03426, 2018

    Leland McInnes, John Healy, and James Melville. Umap: Uniform manifold approximation and 20 Yampolskaya and Mehta projection for dimension reduction.arXiv preprint arXiv:1802.03426, 2018

  31. [40]

    Day, Clint Richardson, Charles K

    Pankaj Mehta, Marin Bukov, Ching-Hao Wang, Alexandre G.R. Day, Clint Richardson, Charles K. Fisher, and David J. Schwab. A high-bias, low-variance introduction to machine learning for physicists.Physics Reports, 810:1–124, May 2019. ISSN 0370-1573. . URL http://dx.doi.org/10.1...

  32. [41]

    Mean-field message-passing equations in the hopfield model and its generaliza- tions.Physical Review E, 95(2):022117, 2017

    Marc M´ ezard. Mean-field message-passing equations in the hopfield model and its generaliza- tions.Physical Review E, 95(2):022117, 2017

  33. [42]

    Transition states and cell fate deci- sions in epigenetic landscapes.Nature Reviews Genetics, 17(11):693–703, 2016

    Naomi Moris, Cristina Pina, and Alfonso Martinez Arias. Transition states and cell fate deci- sions in epigenetic landscapes.Nature Reviews Genetics, 17(11):693–703, 2016

  34. [43]

    Multifarious assembly mixtures: Systems allowing retrieval of diverse stored structures.Proceedings of the National Academy of Sciences, 112(1):54–59, 2015

    Arvind Murugan, Zorana Zeravcic, Michael P Brenner, and Stanislas Leibler. Multifarious assembly mixtures: Systems allowing retrieval of diverse stored structures.Proceedings of the National Academy of Sciences, 112(1):54–59, 2015

  35. [44]

    Stor- age and learning phase transitions in the random-features hopfield model.Physical Review Letters, 131(25):257301, 2023

    Matteo Negri, Clarissa Lauditi, Gabriele Perugini, Carlo Lucibello, and Enrico Malatesta. Stor- age and learning phase transitions in the random-features hopfield model.Physical Review Letters, 131(25):257301, 2023

  36. [45]

    WH Freeman New York, 2004

    Philip Charles Nelson and Philip Nelson.Biological physics. WH Freeman New York, 2004

  37. [46]

    Design principles of 3d epigenetic memory systems.Science, 382(6672):eadg3053, 2023

    Jeremy A Owen, Dino Osmanovi´ c, and Leonid Mirny. Design principles of 3d epigenetic memory systems.Science, 382(6672):eadg3053, 2023

  38. [47]

    Personnaz, I

    L. Personnaz, I. Guyon, and G. Dreyfus. Information storage and retrieval in spin-glass like neural networks.Journal de Physique Lettres, 46(8):359–365, 1985. ISSN 0302-072X. . URL http://dx.doi.org/10.1051/jphyslet:01985004608035900

  39. [48]

    The biological frontier of physics.Physics Today, 59(5): 38–43, 2006

    Rob Phillips and Stephen R Quake. The biological frontier of physics.Physics Today, 59(5): 38–43, 2006

  40. [49]

    Garland Science, 2012

    Rob Phillips, Jane Kondev, Julie Theriot, and Hernan Garcia.Physical biology of the cell. Garland Science, 2012

  41. [50]

    Cellular reprogramming dynamics follow a simple 1d reaction coordinate.Physical Biology, 15(1):016001, 2017

    Sai Teja Pusuluri, Alex H Lang, Pankaj Mehta, and Horacio E Castillo. Cellular reprogramming dynamics follow a simple 1d reaction coordinate.Physical Biology, 15(1):016001, 2017

  42. [51]

    Archishman Raju and Eric D. Siggia. A geometrical perspective on development.Development, Growth & Differentiation, 65(5):245–254, 2023. . URLhttps://onlinelibrary.wiley.com/ doi/abs/10.1111/dgd.12855

  43. [52]

    Hopfield networks is all you need.arXiv preprint arXiv:2008.02217, 2020

    Hubert Ramsauer, Bernhard Sch¨ afl, Johannes Lehner, Philipp Seidl, Michael Widrich, Thomas Adler, Lukas Gruber, Markus Holzleitner, Milena Pavlovi´ c, Geir Kjetil Sandve, et al. Hopfield networks is all you need.arXiv preprint arXiv:2008.02217, 2020

  44. [53]

    Geom- etry of gene regulatory dynamics.Proceedings of the National Academy of Sciences, 118(38): e2109729118, 2021

    David A Rand, Archishman Raju, Meritxell S´ aez, Francis Corson, and Eric D Siggia. Geom- etry of gene regulatory dynamics.Proceedings of the National Academy of Sciences, 118(38): e2109729118, 2021

  45. [54]

    Daydreaming hopfield networks and their surprising effectiveness on correlated data.Neural Networks, 186:107216, June 2025

    Ludovica Serricchio, Dario Bocchi, Claudio Chilin, Raffaele Marino, Matteo Negri, Chiara Cammarota, and Federico Ricci-Tersenghi. Daydreaming hopfield networks and their surprising effectiveness on correlated data.Neural Networks, 186:107216, June 2025. ISSN 0893-6080. . URLht...

  46. [55]

    Emergent properties of collective gene-expression patterns in multicellular systems.Cell Reports Physical Science, 4(2), 2023

    Matthew Smart and Anton Zilman. Emergent properties of collective gene-expression patterns in multicellular systems.Cell Reports Physical Science, 4(2), 2023

  47. [56]

    Information processing in dynamical systems: Foundations of harmony theory

    Paul Smolensky. Information processing in dynamical systems: Foundations of harmony theory. 27:194–281, 1986

  48. [57]

    Cell cycle time series gene expression data encoded as cyclic attractors in hopfield systems

    Anthony Szedlak, Spencer Sims, Nicholas Smith, Giovanni Paternostro, and Carlo Piermaroc- chi. Cell cycle time series gene expression data encoded as cyclic attractors in hopfield systems. PLoS computational biology, 13(11):e1005849, 2017

  49. [58]

    Modeling the attractor landscape of disease progres- sion: a network-based approach.Frontiers in genetics, 8:48, 2017

    Atefeh Taherian Fard and Mark A Ragan. Modeling the attractor landscape of disease progres- sion: a network-based approach.Frontiers in genetics, 8:48, 2017

  50. [59]

    Visualizing data using t-sne.Journal of machine learning research, 9(11), 2008

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of machine learning research, 9(11), 2008. www.annualreviews.org • Hopfield Models 21

  51. [60]

    Van Hemmen, L.B

    J.L. Van Hemmen, L.B. Ioffe, R. K¨ uhn, and M. Vaas. Increasing the efficiency of a neural network through unlearning.Physica A: Statistical Mechanics and its Applications, 163(1): 386–392, February 1990. ISSN 0378-4371. . URLhttp://dx.doi.org/10.1016/0378-4371(90) 90345-S

  52. [61]

    Routledge, 2014

    Conrad Hal Waddington.The strategy of the genes. Routledge, 2014

  53. [62]

    Riemondy, Michael Douglas, Kendall Foster, Nisha Patel, Norihito Kaku, Alexander Linsalata, Jean Nemzek, Brian M

    Fa Wang, Christopher Ting, Kent A. Riemondy, Michael Douglas, Kendall Foster, Nisha Patel, Norihito Kaku, Alexander Linsalata, Jean Nemzek, Brian M. Varisco, Erez Cohen, Jasmine A. Wilson, David W.H. Riches, Elizabeth F. Redente, Diana M. Toivola, Xiaofeng Zhou, Bethany B. Moo...

  54. [63]

    Quantifying the waddington landscape and biological paths for development and differentiation.Proceedings of the National Academy of Sciences, 108(20):8257–8262, 2011

    Jin Wang, Kun Zhang, Li Xu, and Erkang Wang. Quantifying the waddington landscape and biological paths for development and differentiation.Proceedings of the National Academy of Sciences, 108(20):8257–8262, 2011

  55. [64]

    The tolman-eichenbaum machine: unifying space and relational memory through generalization in the hippocampal formation.Cell, 183(5):1249– 1263, 2020

    James CR Whittington, Timothy H Muller, Shirley Mark, Guifen Chen, Caswell Barry, Neil Burgess, and Timothy EJ Behrens. The tolman-eichenbaum machine: unifying space and relational memory through generalization in the hippocampal formation.Cell, 183(5):1249– 1263, 2020

  56. [65]

    Relating transform- ers to models and neural representations of the hippocampal formation.arXiv preprint arXiv:2112.04035, 2021

    James CR Whittington, Joseph Warren, and Timothy EJ Behrens. Relating transform- ers to models and neural representations of the hippocampal formation.arXiv preprint arXiv:2112.04035, 2021

  57. [66]

    How to build a cognitive map.Nature neuroscience, 25(10):1257–1272, 2022

    James CR Whittington, David McCaffary, Jacob JW Bakermans, and Timothy EJ Behrens. How to build a cognitive map.Nature neuroscience, 25(10):1257–1272, 2022

  58. [67]

    sctop: physics-inspired order parameters for cellular identification and visualization

    Maria Yampolskaya, Michael J Herriges, Laertis Ikonomou, Darrell N Kotton, and Pankaj Mehta. sctop: physics-inspired order parameters for cellular identification and visualization. Development, 150(21):dev201873, 2023

  59. [68]

    Finding signatures of low- dimensional geometric landscapes in high-dimensional cell fate transitions.bioRxiv, 2025

    Maria Yampolskaya, Laertis Ikonomou, and Pankaj Mehta. Finding signatures of low- dimensional geometric landscapes in high-dimensional cell fate transitions.bioRxiv, 2025. . URLhttps://www.biorxiv.org/content/early/2025/05/30/2025.05.27.656406

  60. [69]

    Associative pattern recognition through macro-molecular self-assembly.Journal of Statistical Physics, 167:806–826, 2017

    Weishun Zhong, David J Schwab, and Arvind Murugan. Associative pattern recognition through macro-molecular self-assembly.Journal of Statistical Physics, 167:806–826, 2017. 22 Yampolskaya and Mehta

  61. [2023]

    URLhttps://arxiv.org/abs/2304.14964

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.