Pith. sign in

REVIEW 4 major objections 4 minor 72 references

Criticality of a Stochastic Dense Associative Memory Model with Exponential Interaction Function

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Adding random pixel flips to an exponential associative-memory network produces a sharp noise-driven critical transition in which recall drops and the overlap dynamics acquire persistent long-range memory (H≈1.3).

desk verdict Plausible Q transition and a cleanly specified stochastic modern Hopfield model, but the H≈1.3 long-memory claim rests on DFA fits that are not separated from trend artifacts. read the letter →

arxiv 2509.17152 v2 pith:3JIWAJIS submitted 2025-09-21 physics.app-ph nlin.CDphysics.comp-ph

classification physics.app-phnlin.CDphysics.comp-ph
keywords denseassociativememoryexponentialinteractionfunctioncriticalityphasetransitionsalt-and-peppernoisedetrendedfluctuationanalysisanomalousdiffusionMNIST
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 paper claims that a stochastic associative-memory network with an exponential interaction function—a design with very high storage capacity—undergoes a genuine noise-driven phase transition when trained on realistic correlated images (MNIST). Taking the salt-and-pepper flip probability p as the control parameter, the time-averaged recall overlap Q drops sharply at a critical value p_c that falls as the number of stored patterns grows, from about 0.3 at low load to about 0.23 at N=10000. At the same p_c, the overlap time series changes character: below and above the transition its fluctuations are indistinguishable from normal Brownian motion (H≈0.5), while in the critical window they become strongly persistent, with anomalous diffusion exponent H≈1.3 on MNIST and about 1.5 on random Rademacher patterns. If right, this establishes that high-capacity associative memories have a finite-noise critical region in which the network keeps a long, non-Markovian temporal memory, and that this criticality is governed by the interplay of stored-pattern structure and noise rather than by thermal equilibrium.

What carries the argument

The load-bearing object is the stochastic update rule of Eq. (11): the state of each neuron at time t is the sign of a difference between two sums of exponentials of overlap with all stored patterns, and this sign is multiplied by an independent dichotomous random variable that flips the neuron with probability p. The exponential interaction function F(z)=exp(z) inside the energy E=−Σ_i F(x_i^T S) is what gives the model its exponential storage capacity; the multiplicative salt-and-pepper noise turns retrieval into a stochastic process in the overlap Q_t. Two order parameters are then read off Q_t: the time average Q (one-time statistics, recall quality) and the DFA scaling exponent H (two-time statistics, temporal memory), defined through G(t)∼t^H on the integrated overlap X_t. The claimed mechanism is that stored-pattern connectivity and noise compete: at low p the network sticks near one energy well (retrieval but Markovian fluctuations), at high p noise dominates (random walk), and only in between do the residual fluctuations develop long-range correlations.

What would settle it

Take the N=1000 run at p≈p_c, split the 200,000-step overlap series into two independent halves, and re-estimate H from each half over a range of detrended-fluctuation-analysis window sizes; if the two halves disagree by more than the reported uncertainty, or if the local log–log slope drifts monotonically with window length instead of plateauing, then the self-similarity assumption fails and H≈1.3 cannot be interpreted as a well-defined exponent.

Watch

Extended reading notes

Core claim

In the paper's own terms, the central result is the emergence of a critical transition in both the time-averaged overlap Q and the diffusion scaling H, with the critical noise level p_c decreasing as the load N increases. For each load, H marks a transition between a sub-critical and a super-critical regime, both with short-range correlated dynamics, while in the critical regime, found in the range p_c ≈ 0.23–0.3, the MNIST-based network displays long-range correlated dynamics with highly persistent temporal memory, H ≈ 1.3 (the abstract quotes 1.25). Rademacher-trained networks show the same qualitative behavior with a slightly higher memory index H ≈ 1.5. The paper presents Q and H as two order parameters and emphasizes that no equilibrium condition is assumed: the system is run out of equilibrium with a multiplicative noise that acts like random pixel errors.

Load-bearing premise

The central claim rests on the overlap signal having one consistent scaling law at long times; if the critical-region signal shifts its character over the run, the measured H≈1.3 would be an artifact rather than a real exponent.

Editorial extensions

If this is right

  • Storage capacity and noise tolerance are dynamically coupled: p_c drops steeply from about 0.3 at N=5 to about 0.25 at N=100, then slowly to about 0.23 at N=10000, so larger memories tolerate less pixel-flip noise before losing recall.
  • In the critical window the overlap time series is non-Markovian and persistent (H≈1.3), meaning retrieval errors are not independent events but carry long temporal correlations.
  • Sub-critical and super-critical phases both show H≈0.5 Brownian, Markovian fluctuation dynamics, so the transition's special character appears only at criticality rather than as a gradual change in memory.
  • The same p_c appears in both order parameters Q and H, giving a clean dynamical signature of criticality in an out-of-equilibrium associative memory.
  • Pattern correlations matter: MNIST (correlated patterns) yields H≈1.3 while Rademacher (uncorrelated patterns) yields H≈1.5, showing the long-memory exponent is sensitive to the statistical structure of stored items.

Reading between the lines

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

  • If the reported H≈1.3 genuinely reflects a power-law scaling, then the overlap time series at p_c should show a 1/f-type power spectrum and algebraically decaying autocorrelation; measuring those directly on the same simulations would be a sharper check than DFA alone.
  • A testable extension the paper does not perform is an adiabatic sweep of p upward and downward: because the H-jump looks first-order-like, hysteresis in Q or H would indicate a discontinuous transition with practical consequences for noise-tolerant retrieval.
  • The Q-and-H protocol is a new observable for associative-memory criticality and could be applied to other interaction functions (polynomial n-body, softmax attention) to see whether long-range temporal memory at criticality is unique to the exponential choice or generic to high-capacity memories.
  • If the critical long-memory regime has a functional role, a network deliberately operated at p_c could serve as a temporal-memory buffer or a source of slow fluctuations; the authors do not claim this, but it is a plausible engineering consequence.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper studies a stochastic version of the exponential modern Hopfield network (called SMHN in the body and SEDAM in the abstract) with multiplicative salt-and-pepper noise, trained on MNIST patterns. It defines the time-averaged overlap Q and the detrended fluctuation analysis (DFA) exponent H as order parameters and reports a noise-driven transition in both, with the critical noise level p_c decreasing with the number of stored patterns N. The central claim is that at critical noise levels near p_c≈0.23–0.3 the network enters a long-range correlated dynamical regime with H≈1.3, interpreted as highly persistent temporal memory. The model and simulation protocol are explicit, but the evidence for the H transition and for a genuine phase transition is not yet sufficient.

Significance. If the claims were established, this would be a valuable contribution: it would extend the phase-diagram literature of modern Hopfield networks from equilibrium thermodynamic analysis to out-of-equilibrium, noise-driven temporal dynamics, and it would demonstrate a load-dependent critical noise level in a realistic correlated dataset (MNIST). The manuscript is explicit about the model, the simulation parameters (L=784, 200,000 time steps, 25 noise values, loads up to 10,000), and it is methodologically interesting to use DFA on the overlap time series. The main novelty, however, is the claimed critical-regime exponent H≈1.3, and that result currently rests on an untested monoscaling assumption and is vulnerable to DFA trend artifacts. The comparison with Rademacher patterns and with a stochastic Hopfield network is announced in the abstract but is not actually reported in the body, which weakens the paper's stated scope.

major comments (4)
  1. [Overlap as order parameter, Eqs. (13)–(15) and End Matter, Eq. (17)] The reported H≈1.3 in the critical regime is not supported because H>1 cannot arise from stationary increments of a bounded signal. Since |Q_t|≤1, the integrated variable satisfies |X(t+Δt)-X(t)|≤Δt, so any stationary-increment process has F(Δt) ≤ C Δt and hence H≤1. A fitted H≈1.3 therefore indicates a nonstationary mean or a local trend that the linear detrending of Eq. (17) does not remove; a long dwell in one energy well followed by a switch creates quadratic segments in X_t, which DFA-1 can mistake for superdiffusion. The manuscript provides no stationarity test, no surrogate data, no DFA-order variation, and no trajectory-length convergence check to rule out this artifact.
  2. [Results, Fig. 3 and accompanying text] The authors state that the DFA function in the critical region 'always' shows a crossover between short- and long-time scaling, yet Eq. (14) defines H through a single monoscaling exponent. When a crossover is present, the reported 'H≈1.3' is not a well-defined scaling exponent unless the two fitting ranges are justified and the fits are shown to be stable; the paper does not report the fitting ranges, the goodness of fit, or how the short- and long-time regimes are separated. This is load-bearing because the long-memory claim is precisely the claim of a single long-range scaling regime.
  3. [Abstract and body (comparison claims)] The abstract states that the MNIST-based SEDAM is compared with a SEDAM trained on Rademacher patterns and with a stochastic Hopfield network, and that the Rademacher-trained network gives H≈1.5. However, no section, figure, or table in the body reports these simulations, their parameter settings, or the comparison results. As written, the abstract's comparison claim is unsupported by the presented material.
  4. [Results, Figs. 1–4; 'phase transition' language] The 'phase transition' in Q and H is inferred from a single realization per (N,p) on a fixed L=784 system with 200,000 time steps; there are no error bars, no averaging over initial conditions or pattern subsets, and no finite-size scaling or thermodynamic-limit analysis. Moreover, p_c is identified from the Q(p) curves and then the same p_c values are used to label the H anomaly in Fig. 4, so the coincidence of the two transitions is partly by construction. A finite-size analysis and an independent identification of p_c are needed before calling the regime 'critical' or describing the transition as first-order.
minor comments (4)
  1. [Notation throughout] The abstract uses 'SEDAM' and 'load K', while the body uses 'SMHN' and 'load N'; please unify the terminology to avoid confusion.
  2. [Results, Fig. 2 caption] The paper should state how p_c is estimated from the Q(p) curves (e.g., threshold criterion, inflection point, or midpoint between high- and low-Q plateaus), because the reported p_c values are the basis for the subsequent H labeling.
  3. [End Matter, Eq. (15)] The notation G^2(t)=σ^2(t) reuses G(t) from Eq. (14) in a way that is potentially confusing; please define G(t) and σ^2(t) separately and consistently.
  4. [Discussion section] There are several typos and stylistic slips, including 'Browian' for 'Brownian' and 'a one-time statistical average of of the overlap parameter'; these should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported H≈1.3 critical anomaly is an independent DFA measurement, not encoded in the model or derived from the Q-based p_c calibration.

full rationale

The derivation chain is self-contained. The stochastic exponential DAM update rule (Eq. 11) defines the dynamics; the two order parameters are measured from the simulated overlap time series: Q is the time average of Q_t (Eq. 13), and H is the DFA scaling exponent of the integrated overlap X_t (Eqs. 14–20). Nothing in these definitions or in the model equations encodes the reported critical values p_c≈0.23–0.3 or the anomalous exponent H≈1.3, so the central claim does not reduce to its inputs by construction. The only dependency is that p_c is identified from the Q(p) curves, and the same p_c values are used to label the H regime in Figs. 3–4; this is a calibration/selection coupling between the two order parameters, but the H values themselves are independently measured and could in principle have been H≈0.5 at those points. Self-citations (e.g., refs. 25–28 and 68–69) support the standard DFA methodology and prior applications, not the target result, so they are not load-bearing. The paper itself states a limitation: 'The well-posedness of H in Eq. (14) is based on the assumption that X_t is a self-similar, i.e., monoscaling signal,' and it notes that 'this crossover is always seen in the critical region,' which weakens the interpretation of a single H but is a statistical-validity concern, not a circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The model uses a known exponential energy function; the load-bearing numbers are empirical estimates p_c and H. No new physical entities are introduced. The leading unstated premises are the DFA monoscaling assumption, the unreported MNIST binarization, and the use of time averages for a nonstationary overlap signal.

free parameters (2)
  • Critical noise threshold p_c = 0.23-0.30 depending on N
    Extracted from the p-dependence of the time-averaged overlap Q and the DFA discontinuity; it is the central threshold of the claimed transition and is not predicted by a theory in the paper.
  • DFA scaling exponent H = about 0.5 in sub/super-critical phases; about 1.3 at criticality
    Obtained by linear least-squares fit of log F(dt) versus log dt via DFA; it is the central evidence for persistent temporal memory, and no confidence intervals are reported.
assumptions (4)
  • domain assumption The integrated overlap X_t is self-similar and monoscaling, so G(t) approximately t^H gives a well-posed H.
    Invoked in the 'Overlap as order parameter' section before Eq. (14); if X_t is multifractal or has multiple crossover scales, H is not a single robust exponent.
  • domain assumption MNIST images are converted to binary plus-or-minus-one patterns with a thresholding procedure that is not reported.
    The simulation setup stores 'binary images' with K=784 pixels; the binarization affects inter- and intra-pattern correlations and hence the measured p_c, but the threshold is unstated.
  • domain assumption The multiplicative noise xi_t[l] is independent across time and neurons, with probability of negative sign equal to p.
    Stated in the model definition around Eq. (11); this noise model, not thermal equilibrium, defines the dynamics and the control parameter.
  • domain assumption The time-averaged overlap Q in Eq. (13) is a meaningful order parameter for a non-equilibrium, possibly nonstationary process.
    Used throughout the Results; if Q_t is nonergodic or has divergent correlation time, a single 200,000-step time average may not represent the phase.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Criticality of a Stochastic Dense Associative Memory Model with Exponential Interaction Function." pith.science (2026). https://pith.science/paper/3JIWAJIS

@misc{pith2026250917152,
  author       = {Pith},
  title        = {Pith review of: Criticality of a Stochastic Dense Associative Memory Model with Exponential Interaction Function},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3JIWAJIS}},
  note         = {Machine review of arXiv:2509.17152}
}
read the original abstract

The Hopfield network (HN) is a classical model of associative memory with stored patterns encoded as minima of an energy function shaped by a Hebbian learning rule. Dense Associative Memory (DAM) models introduce n-body interactions among neurons with n greater than 2 and, more recently, also exponential interaction functions, which significantly improve the network's storing capacity. While the emergence of phase transitions in HN and DAM were extensively studied, the investigation of exponential DAM is still in its early stages. Further, an equilibrium thermodynamical condition is typically assumed, while out-of-equilibrium dynamics are not considered. Here, we study the temporal dynamics of a stochastic exponential DAM (SEDAM) with a multiplicative salt-and-pepper noise and trained on the MNIST dataset. While taking the noise probability p as control parameter, the time-averaged overlap Q and the diffusion scaling H are taken as order parameters, being H related to the network's time correlation features. The MNIST-based SEDAM is also compared with a SEDAM trained on standard Rademacher patterns and with a stochastic HN (SHN). We found the emergence of a phase transition in both Q and H, with the critical noise level pc decreasing as the load K increases. For each load K, Q highlights a transition between a sub-critical and a super-critical regime, both with short-time correlated dynamics. Conversely, in the critical regime of the MNIST-based SEDAM the network displays long-time correlated dynamics with highly persistent temporal memory marked by the high value H around 1.25. Similar behaviors are observed for both models trained with Rademacher patterns, but with a slightly higher temporal memory index H around 1.5.

Figures

Figures reproduced from arXiv: 2509.17152 by the authors.

Figure 1
Figure 1. FIG. 1: Average overlap parameter [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Critical noise levels ( [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3: DFA curves for networks with [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: FIG. 4: Second moment scaling parameter ( [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 60 canonical work pages

  1. [1]

    catastrophic forgetting

    Consequently, the natural extension of the model, introduced by Amit and co-workers, is to consider the mathematical formalism of the canonical ensemble in the framework of statisti- cal mechanics [18], i.e., to consider the Hopfield network in thermal equilibrium with a thermal bath at a temperatureT >0, which is equiva- lent to introducing randomness in...

  2. [2]

    Hopfield, Proceedings of the National Academy of Sciences of the United States of America79, 2554 – 2558 (1982)

    J. Hopfield, Proceedings of the National Academy of Sciences of the United States of America79, 2554 – 2558 (1982)

  3. [3]

    McCulloch and Pitts neurons usually assume values±1 or{0,1}, but whatever couple of val- ues{a, b}could be used

  4. [4]

    D. O. Hebb,The Organization of Behavior: A Neuropsychological Theory(Wiley & Sons, New York, 1949)

  5. [5]

    no firing

    Also the valuex i = 0 is often used to define the “no firing” state of thei-th neuron

  6. [6]

    On the contrary, we here limit to show the case of synchronous update given in Eq

    In the original Hopfield proposal [1] and in many successive studies, the update dynam- ics are taken asynchronous, implying that each neuron randomly update its state in a given time window. On the contrary, we here limit to show the case of synchronous update given in Eq. (3)

  7. [7]

    D. J. Amit, H. Gutfreund, and H. Sompolinsky, Physical Review A32, 1007 – 1018 (1985)

  8. [8]

    Forrest, Journal of Physics A21, 245 – 255 (1988)

    B. Forrest, Journal of Physics A21, 245 – 255 (1988)

Show all 72 references
  1. [9]

    Hopfield, Proceedings of the National Academy of Sciences of the United States of America81, 3088 – 3092 (1984)

    J. Hopfield, Proceedings of the National Academy of Sciences of the United States of America81, 3088 – 3092 (1984)

  2. [10]

    Clearly, the distorted pattern, which is a noisy version of the stored pattern, must belong to the attraction basin of the stored pattern itself. A weak, and thus broader, definition of “good” retrieval is often exploited and involves the in- clusion of an estimation error, e....

  3. [11]

    Krotov and J

    D. Krotov and J. J. Hopfield, inProceed- ings of the 30th International Conference on Neural Information Processing Systems (NIPS2016), Advances in Neural Information Processing Systems, Vol. 29, edited by D. Lee, M. Sugiyama, U. Luxburg, I. Guyon, and R. Garnett (Curran Assoc...

  4. [12]

    Gardner, Nuclear Physics, Section B257, 747 – 765 (1985)

    E. Gardner, Nuclear Physics, Section B257, 747 – 765 (1985)

  5. [13]

    Gardner, Journal of Physics A: Mathemati- cal and General20, 3453 – 3464 (1987)

    E. Gardner, Journal of Physics A: Mathemati- cal and General20, 3453 – 3464 (1987)

  6. [14]

    The main difference between the model studied by Gardner and that proposed by Krotov and Hopfield lies in the dynamical update rule, the first one being based on the magnetic field and the second one on the energy

  7. [15]

    Demircigil, J

    M. Demircigil, J. Heusel, M. L¨ owe, S. Upgang, and F. Vermet, Journal of Statistical Physics 168, 288 – 299 (2017)

  8. [16]

    These authors also proved the claim of Krotov and Hopfield 2016 [10] regarding the storage capacity of the MHN with n-body interactions

  9. [17]

    V. E. Ising, Zeitschrift f¨ ur Physik31, 253 (1925)

  10. [18]

    Parisi, Physical Review Letters43, 1754 (1979)

    G. Parisi, Physical Review Letters43, 1754 (1979)

  11. [19]

    Huang,Statistical Mechanics, 2nd ed

    K. Huang,Statistical Mechanics, 2nd ed. (John Wiley & Sons, 1987)

  12. [20]

    D. J. Amit and V. Martin-Mayor,Field theory, the renormalization group, and critical phe- nomena: Graphs to computers: Third edition, 3rd ed. (World Scientific, 2005) p. 1 – 544

  13. [21]

    Cardy,Scaling and renormalization in sta- tistical physics(Cambridge University Press,

    J. Cardy,Scaling and renormalization in sta- tistical physics(Cambridge University Press,

  14. [22]

    Grigolini, Chaos Solit

    P. Grigolini, Chaos Solit. Fractals81, 575 (2015)

  15. [23]

    Hunt von Herbing, L

    I. Hunt von Herbing, L. Tonello, M. Ben- fatto, A. Pease, and P. Grigolini, Entropy23, 10.3390/e23091141 (2021)

  16. [24]

    J. M. Beggs and D. Plenz, Journal of neuro- science23, 11167 (2003)

  17. [25]

    J. M. Beggs and D. Plenz, Journal of neuro- science24, 5216 (2004)

  18. [26]

    Allegrini, P

    P. Allegrini, P. Paradisi, D. Menicucci, and A. Gemignani, Front. Physio.1, 10.3389/fphys.2010.000128 (2010), dOI: 10.3389/fphys.2010.000128

  19. [27]

    Paradisi, P

    P. Paradisi, P. Allegrini, A. Gemignani, 11 M. Laurino, D. Menicucci, and A. Piarulli, AIP Conf. Proc1510, 151 (2013), dOI: 10.1063/1.4776519

  20. [28]

    Allegrini, P

    P. Allegrini, P. Paradisi, D. Menicucci, M. Lau- rino, R. Bedini, A. Piarulli, and A. Gemignani, Chaos Solitons Fract.55, 32 (2013)

  21. [29]

    Allegrini, P

    P. Allegrini, P. Paradisi, D. Menicucci, M. Lau- rino, A. Piarulli, and A. Gemignani, Phys. Rev. E Stat. Nonlin. Soft Matter Phys92, 10.1103/PhysRevE.92.032808 (2015)

  22. [30]

    Deco and V

    G. Deco and V. K. Jirsa, Journal of Neuro- science32, 3366 (2012)

  23. [31]

    Scarpetta and A

    S. Scarpetta and A. de Candia, PLOS ONE8, 1 (2013)

  24. [32]

    Sorrentino, E

    P. Sorrentino, E. Troisi Lopez, A. Romano, C. Granata, M. C. Corsi, G. Sorrentino, and V. Jirsa, NeuroImage277, 120260 (2023)

  25. [33]

    Zimmern, Frontiers in Neural Circuits14, 10.3389/fncir.2020.00054 (2020)

    V. Zimmern, Frontiers in Neural Circuits14, 10.3389/fncir.2020.00054 (2020)

  26. [34]

    Turalska, B

    M. Turalska, B. J. West, and P. Grigolini, Physical Review E83, 061142 (2011)

  27. [35]

    Mahmoodi, B

    K. Mahmoodi, B. J. West, and P. Grigolini, Complexity2018, 8139058 (2018)

  28. [36]

    Bertschinger and T

    N. Bertschinger and T. Natschl¨ ager, Neural Computation16, 1413 (2004)

  29. [37]

    Boedecker, O

    J. Boedecker, O. Obst, J. T. Lizier, N. M. Mayer, and M. Asada, Theory in Biosciences 131, 10.1007/s12064-011-0146-8 (2012)

  30. [38]

    S. S. Schoenholz, J. Gilmer, S. Ganguli, and J. Sohl-Dickstein, Deep information propaga- tion (2017), arXiv:1611.01232

  31. [39]

    Shriki and D

    O. Shriki and D. Yellin, PLOS Comp. Bio.12, 1 (2016)

  32. [40]

    Haldeman and J

    C. Haldeman and J. M. Beggs, Phys. Rev. Lett. 94, 058101 (2005)

  33. [41]

    Maass, T

    W. Maass, T. Natschl¨ ager, and H. Markram, Neural Computation14, 2531 (2002)

  34. [42]

    Fraiman, P

    D. Fraiman, P. Balenzuela, J. Foss, and D. R. Chialvo, Physical Review E79, 061922 (2009), dOI: 10.1103/PhysRevE.79.061922

  35. [43]

    Chialvo, Nature Physics6, 744 (2010), dOI: 10.1038/nphys1803

    D. Chialvo, Nature Physics6, 744 (2010), dOI: 10.1038/nphys1803

  36. [44]

    B. West, E. Geneston, and P. Grigolini, Phys. Rep.468, 1 (2008), dOI: 10.1016/j.physrep.2008.06.003

  37. [45]

    Vanni, M

    F. Vanni, M. Lukovi´ c, and P. Grigolini, Phys. Rev. Lett.107, 10.1103/Phys- RevLett.107.078103 (2011), dOI: 10.1103/PhysRevLett.107.078103

  38. [46]

    Angiolelli, S

    M. Angiolelli, S. Scarpetta, P. Sorrentino, E. T. Lopez, M. Quarantelli, C. Granata, G. Sor- rentino, V. Palmieri, G. Messuti, M. Ste- fano, S. Filippi, C. Cherubini, A. Loppini, and L. Chiodo, bioRxiv 10.1101/2024.07.15.603226 (2024)

  39. [47]

    In analogy with THN, we denote as TMHN the thermodynamical version of MHN, in particu- lar n-body TMHN and exponential TMHN

  40. [48]

    D. J. Amit, H. Gutfreund, and H. Sompolin- sky, Physical Review Letters55, 1530 – 1533 (1985)

  41. [49]

    D. J. Amit, H. Gutfreund, and H. Sompolinsky, Annals of Physics173, 30 – 67 (1987)

  42. [50]

    Lucibello and M

    C. Lucibello and M. M´ ezard, Physical Review Letters132, 10.1103/PhysRevLett.132.077301 (2024)

  43. [51]

    D. J. Amit, H. Gutfreund, and H. Sompolinsky, Physical Review A35, 2293 – 2303 (1987)

  44. [52]

    Gardner, Europhysics Letters4, 481 – 485 (1987)

    E. Gardner, Europhysics Letters4, 481 – 485 (1987)

  45. [53]

    Ramsauer, B

    H. Ramsauer, B. Sch¨ afl, J. Lehner, P. Seidl, M. Widrich, T. Adler, L. Gruber, M. Hol- zleitner, D. Kreil, M. Kopp, G. Klam- bauer, J. Brandstetter, and S. Hochreiter, inICLR 2021 - 9th International Conference on Learning Representations(2021) preprint: arXiv:2008.02217v3

  46. [54]

    As an example, this noise term can also be in- terpreted as a disturbance continuously acting on the pixels of a process of image retrieval

  47. [55]

    Deng, IEEE Signal Processing Magazine29, 141 (2012)

    L. Deng, IEEE Signal Processing Magazine29, 141 (2012)

  48. [56]

    This is often denoted as Hurst exponent, even if it corresponds to the exponent introduced by Hurst in his milestone paper [67] only in the case of a monoscaling signal

  49. [57]

    C.-K. Peng, S. V. Buldyrev, S. Havlin, M. Si- mons, H. E. Stanley, and A. L. Goldberger, Phys. Rev. E49, 1685 (1994)

  50. [58]

    C.-K. Peng, S. Havlin, H. E. Stanley, and A. L. Goldberger, Chaos5, 82 – 87 (1995)

  51. [59]

    Stanley, L

    H. Stanley, L. Amaral, A. Goldberger, S. Havlin, P. Ivanov, and C.-K. Peng, Physica A270, 309 – 324 (1999)

  52. [60]

    J. W. Kantelhardt, E. Koscielny-Bunde, H. H. Rego, S. Havlin, and A. Bunde, Physica A295, 441 – 454 (2001)

  53. [61]

    M. V. Tsodyks and M. V. Feigel’man, Euro- physics Letters6, 101 (1988)

  54. [62]

    Shiino and T

    M. Shiino and T. Fukai, Journal of Physics A: Mathematical and General25, L375 (1992)

  55. [63]

    De Marzo and G

    G. De Marzo and G. Iannelli, Physica A612, 10.1016/j.physa.2023.128487 (2023)

  56. [64]

    Negri, C

    M. Negri, C. Lauditi, G. Perugini, C. Lucibello, and E. Malatesta, Physical Review Letters 131, 10.1103/PhysRevLett.131.257301 (2023). 12

  57. [65]

    Serricchio, D

    L. Serricchio, D. Bocchi, C. Chilin, R. Marino, M. Negri, C. Cammarota, and F. Ricci-Tersenghi, Neural Networks186, 10.1016/j.neunet.2025.107216 (2025)

  58. [66]

    K. Musa, S. Kumar, M. Katidis, and Y.-P. Huang, Dense associative memory in a non- linear optical hopfield neural network (2025), arXiv:2506.07849 [physics.optics]

  59. [67]

    dis- torted

    This corresponds to the fixed point related to the stored pattern closest to the initial “dis- torted” patternS t=0 = exi

  60. [68]

    Hurst, Trans

    H. Hurst, Trans. Am. Soc. Civil Eng.116, 770 (1951)

  61. [69]

    Paradisi, R

    P. Paradisi, R. Cesari, A. Donateo, D. Contini, and P. Allegrini, Nonlinear Processes in Geo- physics19, 113 (2012), note: Corrigendum: Paradisi et al., Nonlin. Processes Geophys.19, 685 (2012), doi: 10.5194/npg-19-685-2012

  62. [70]

    Paradisi and P

    P. Paradisi and P. Allegrini, inComplexity and Nonlinearity in Cardiovascular Signals, edited by R. Barbieri, E. P. Scilingo, and G. Valenza (Springer, Cham, 2017) pp. 161–195

  63. [71]

    S. K. Nayak, A. Bit, A. Dey, B. Mohapatra, and K. Pal, Journal of Healthcare Engineering 2018, 10.1155/2018/6920420 (2018)

  64. [72]

    Rydin Gorj˜ ao, G

    L. Rydin Gorj˜ ao, G. Hassan, J. Kurths, and D. Witthaut, Computer Physics Communica- tions273, 108254 (2022). END MA TTER Detrended Fluctuation Analysis Detrended Fluctuation Analysis (DFA) is a widely used method to detect scaling in time series and was proven to be robust w...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.