Building Intuition for Dynamical Mean-Field Theory: A Simple Model and the Cavity Method

Emmy Blumenthal

REVIEW 2 major objections 4 minor 2 cited by

Building Intuition for Dynamical Mean-Field Theory: A Simple Model and the Cavity Method

T0 review · 2 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Dynamical mean-field theory is proven exact for a linear random-coupling model: $N$-body dynamics collapse to one stochastic equation, and the cavity method extends this to the nonlinear Lotka–Volterra model.

desk verdict A clear teaching tutorial whose final DMFT equations are right, but whose Section 3.6 contains a false Laplace-transform identity that needs fixing before it can serve as the exact walkthrough it promises. read the letter →

arxiv 2507.16654 v1 pith:VLCN6RGF submitted 2025-07-22 cond-mat.dis-nn physics.bio-ph

Emmy Blumenthal This is my paper · ORCID

classification cond-mat.dis-nnphysics.bio-ph

keywords dynamicalmean-fieldtheorycavitymethodrandommatrixLotka–Volterramodelself-averagingskew-symmetriccouplingsmemorykerneldisorderedsystems

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

Dynamical mean-field theory (DMFT) replaces a system of many coupled equations of motion with a single stochastic equation for one representative variable, but the replacement is usually a leap of faith. This tutorial makes the leap concrete by solving exactly a linear model with random skew-symmetric couplings: in the large-$N$ limit the many-body dynamics provably reduce to a single integro-differential equation with a Gaussian noise term and a memory kernel, both given by the Bessel function $J_1(2\sigma t)/(\sigma t)$. The derivation reveals precisely which assumption carries the reduction, namely self-averaging of order parameters, and proves it for the linear model by explicit $O(1/N)$ variance estimates. The same cavity method that re-derives the linear result is then applied to the nonlinear generalized Lotka–Volterra model, where self-averaging is asserted rather than proven and the payoff is an effective single-species equation with mean-field, noise, and back-reaction terms.

What carries the argument

The argument runs on three objects. First, the exact integration-out: solving the equations for the $N-1$ unselected particles in terms of the matrix exponential $e^{A_{\backslash\alpha} t}$ turns the selected particle's equation into $dx_\alpha/dt = \Phi_\alpha(t) - \sigma^2\int_0^t dt'\, \nu_\alpha(t-t')x_\alpha(t')$, cleanly separating direct forcing from memory. Second, self-averaging: the paper proves that shuffled averages (which particle is selected) equal quenched averages (over the random couplings and initial conditions) because the variances of the shuffled order parameters are $O(1/N)$, so under one frozen realization $\Phi$ behaves as a Gaussian process. Third, the cavity method: add one new particle, treat its couplings as a $1/\sqrt{N}$ perturbation, use linear response to express the neighbors' dynamics, average over disorder, and close the equations by asserting self-averaging of $m(t)$, $C(t,t')$, and $\nu(t,t')$. The recurring output is the Bessel kernel $J_1(2\sigma t)/(\sigma t)$, which in the linear model is simultaneously the noise covariance, the response function, and the correlation function (up to $\sigma_0^2$).

What would settle it

Run the full GLV dynamics at increasing species number $S$ and measure the quenched variance of the shuffled order parameter, $\mathrm{Var}_Q\big[\tfrac{1}{S}\sum_i N_i(t)N_i(t')\big]$; Eq. (81) requires this variance to vanish like $1/S$, and a non-vanishing plateau would falsify the DMFT closure for that parameter regime. For the linear model, the analogous check is $\mathrm{Var}_Q\big[\tfrac{1}{N}\sum_i x_i(t)x_i(s)\big]$, which Section 2.4 computes as $O(1/N)$, so observing any slower decay would falsify the exactness of Eq. (26).

Watch

Extended reading notes

Core claim

In the $N\to\infty$ limit, a randomly chosen degree of freedom of the linear model obeys the DMFT equation of motion $dx/dt = \Phi(t) - \sigma^2\int_0^t dt'\, \nu(t-t')x(t')$, where $\Phi$ is a Gaussian process with zero mean and covariance $\sigma^2\sigma_0^2 J_1(2\sigma|t-s|)/(\sigma|t-s|)$, and $\nu(t)=J_1(2\sigma t)/(\sigma t)\,\Theta(t)$ (Eq. 26). The central claim is that this reduction is exact: for the linear model the shuffled and quenched averages coincide because the relevant variances are $O(1/N)$, computed explicitly in Section 2.4. The cavity method re-derives the same equation from a self-consistency condition and yields the correlation function $C(t-s)=\sigma_0^2 J_1(2\sigma|t-s|)/(\sigma|t-s|)$, whose oscillations decay as $t^{-3/2}$, and a power spectrum equal to the Wigner semicircle on $[-2\sigma,2\sigma]$. For the generalized Lotka–Volterra model, the same cavity steps produce the effective single-species equation $dN_0/dt = N_0(1-N_0-\mu m(t)-\sigma\eta(t)-\gamma\sigma^2\int_0^t dt'\, \nu(t,t')N_0(t'))$, whose static limit is a truncated Gaussian abundance distribution.

Load-bearing premise

The load-bearing premise is the self-averaging closure asserted in Step 5 of the cavity method (Eq. 81) — that shuffled and quenched averages of the order parameters coincide — which the paper proves for the linear model but merely asserts for the nonlinear Lotka–Volterra model, so the GLV reduction stands or falls on that unproven equality.

Editorial extensions

If this is right

For the linear model the reduction is exact and testable: the correlation function $C(t)=\sigma_0^2 J_1(2\sigma t)/(\sigma t)$ and the semicircle power spectrum are closed-form predictions, and the paper's simulations at $N=128$ match them.
Each degree of freedom shows quasi-periodic oscillations with period $\pi/\sigma$ that decay only as $t^{-3/2}$, so the disorder-averaged dynamics remains coherent for a long time.
The power spectral density vanishes for $|\omega|>2\sigma$, meaning the system acts as a low-pass filter that responds to external driving only below $2\sigma$.
For the GLV model, the DMFT equation reduces many-species dynamics to a single stochastic process whose static limit is a truncated Gaussian abundance distribution and which exhibits phase transitions to chaos and unbounded growth.
The back-reaction (memory) term in the DMFT equation is unavoidable: naive averaging of $\sum_j A_{ij}x_j$ misses the $O(1)$ feedback that the selected particle exerts on all others and receives back, which is exactly what the cavity method captures.

Reading between the lines

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

In the linear model the correlation and response functions coincide, $C(t)=\sigma_0^2\nu(t)$, a fluctuation-dissipation-like identity that likely follows from the skew-symmetry of the couplings; measuring the ratio $C(t)/\nu(t)$ in the GLV model as $\gamma$ departs from $-1$ would quantify how much of the DMFT structure survives when this symmetry is broken.
The paper's $O(1/N)$ variance estimates for the linear model double as finite-size correction formulas; applying the same statistic to the GLV model at increasing $S$ would give a numerical certificate for the closure in Eq. (81) without solving the full DMFT equations.
The exactly solvable linear model is a natural benchmark for numerical DMFT solvers: any iterative scheme that solves the linear self-consistency equations should recover the Bessel kernel and the semicircle spectrum exactly, so a solver that fails on this test is wrong independent of nonlinear effects.

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.

Desk Editor's Note

read the letter

Thanks for the report. I read the full manuscript and the stress-test note. My take: this is a genuinely useful teaching resource, and the final linear-model DMFT equations are correct and verified by simulation. But the cavity-method derivation in Section 3.6 contains a false Laplace-transform identity that breaks the 'exact walkthrough' promise; the final formulas survive, but the printed algebra does not.

What's good: The linear model is worked out with explicit self-averaging calculations, which is rare in this literature. The cavity method presentation is clear, and the comparison with numerics in Figure 5 is a nice confirmation. The GLV application is honest—it cites Ref. [3] and explicitly labels the self-averaging closure as an assumption, which is the right level of caution for a tutorial. The paper will likely help many biophysics students get past the barrier-to-entry to DMFT.

Where it's soft: The stress-test note is correct. Eq. (56) claims the double Laplace transform of f(t-s) is f̃(z_t)f̃(z_s)/(z_t+z_s); the correct expression is [f̃(z_t)+f̃(z_s)]/(z_t+z_s). Try f(u)=e^{-u} with z_t=1, z_s=2: the paper's formula gives 1/18, the correct one 5/12. Eq. (55) uses the bad identity to go from Eq. (54) to a compact expression. The striking part is that the final C(t-s)=σ0²G(|t-s|) follows as well from the correct identity, so the paper's headline result is not wrong. But a reader who follows the algebra in Section 3.6 cannot reproduce the steps, and the tutorial's purpose is exactly to give a reliable derivation. This is a moderate flaw, not a fatal one.

Another soft spot, which the author acknowledges, is the GLV self-averaging closure. For a tutorial this is acceptable, but the distinction between what is proven for the linear model and what is assumed for GLV could be drawn more sharply.

Who it's for: graduate students and researchers new to DMFT, especially in ecology and biophysics. It deserves a serious referee after the algebra in Section 3.6 is corrected. My recommendation: engage with it, ask for a revision fixing Eq. (56) and providing a corrected derivation of Eq. (55).

Referee Report

2 major / 4 minor

Summary. This tutorial paper introduces dynamical mean-field theory (DMFT) through a deliberately simple linear model with random skew-symmetric couplings, derives the DMFT equation of motion exactly, then presents the cavity method as a general tool and applies it to the generalized Lotka-Volterra model. The central claim is that, in the large-N limit, a randomly chosen degree of freedom obeys a single-particle stochastic integro-differential equation whose noise is Gaussian with a Bessel-function covariance and whose memory kernel is a Bessel-function response. The paper validates the prediction against numerical simulation in Figure 5 and closes with applications to correlations, response, and power spectra.

Significance. If the derivation is corrected, the paper would be a genuinely useful pedagogical resource: it makes explicit the distinction between quenched and 'shuffling' randomness, proves self-averaging in the linear model by computing O(1/N) variances, and provides a step-by-step cavity derivation that reproduces the exact result. The GLV section also gives a concise, well-cited introduction to a widely used DMFT reduction in ecology. The main strength is the exact linear-model treatment with a numerical check; the main weakness is an algebraic error in the cavity-method walkthrough that, although it does not change the final answer, invalidates the derivation as printed and teaches an incorrect general identity.

major comments (2)

[Section 3.6, Eqs. (55)-(56)] The double-Laplace-transform identity in Eq. (56) is false. For a function f(|t-s|), the correct identity is the Laplace transform of f(|t-s|) equals [\tilde f(z_t)+\tilde f(z_s)]/(z_t+z_s), not \tilde f(z_t)\tilde f(z_s)/(z_t+z_s). The paper's Eq. (55) is a consequence of this incorrect identity; using the correct identity and Eq. (48) gives \tilde C(z_t,z_s)=\sigma_0^2[\tilde G(z_t)+\tilde G(z_s)]/(z_t+z_s), which still leads to the correct Eq. (57). As printed, the derivation of Eq. (57) is invalid and the stated 'general' identity is wrong. This is load-bearing because the tutorial's exact walkthrough of the cavity method culminates in this step, and a reader following the algebra cannot reproduce the claimed intermediate result. The fix is local but essential: correct Eq. (55), remove or replace Eq. (56), and show the derivation leading to Eq. (57) with the correct identity.
[Section 5.2.5, Eq. (81)] The GLV cavity derivation closes the DMFT equations by asserting that the empirical order parameters m(t), C(t,t'), and \nu(t,t') are self-averaging, equating shuffled averages with quenched averages. The paper explicitly labels this an assertion and notes that such self-averaging is 'not possible' to prove for most systems. This is acceptable as a stated assumption, but because the entire GLV DMFT reduction depends on it, the text should more prominently flag it as a nontrivial hypothesis and point the reader to the specific justification (e.g., in Ref. [3]) rather than presenting it as a routine step. As written, a reader may not appreciate that the GLV DMFT equation (79) is not derived to the same standard of rigor as the linear-model equation (26).

minor comments (4)

[Section 2.4.3, Eq. (23)] The displayed expression for \langle\nu_\alpha(t)\rangle_Q has a trace index typo: the summand should be [e^{A_{\backslash\alpha}t}]_{ii} rather than [e^{A_{\backslash\alpha}t}]_{ij}.
[Section 3.6, Eq. (59)] The phrase 'shuffling disorder' in the parenthetical describing the Gaussian process should read 'shuffling randomness' for consistency with the terminology introduced in Section 2.2.
[Section 3.7] The statement that A_{0i} and x_{i\backslash 0}(t) are independent 'by definition' is a bit terse; it would be clearer to say that A_{0i} is independent of the submatrix A_{\backslash 0} and of the initial conditions of the other degrees of freedom, which makes the factorization in Eq. (34) valid.
[Section 5.2.6, Eq. (82)] The sentence following Eq. (82) says 'we can see that \nu(t,t')\ge 0', but the displayed ODE does not by itself make positivity obvious; a short justification or reference would help the reader.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the linear-model DMFT derivation is self-contained, with no fitted parameters, no self-citation load-bearing steps, and no definitional reduction of output to input.

full rationale

This tutorial contains no fitted parameters, no self-citation chains, and no definitional identification of output with input. The central linear-model DMFT equation (26) is derived by exactly integrating out N-1 degrees of freedom (Eqs. 9-12), computing the noise covariance from the quenched spectral density (Eqs. 13-18), and proving self-averaging by explicit O(1/N) variance estimates (Eqs. 20-24). The cavity method later re-derives the same result under an explicit self-averaging assertion (Eq. 41), which the paper itself labels as an assumption rather than a proved input; an unproven closure assumption is a limitation, not circularity. The GLV section explicitly follows Ref. [3] as external prior work and does not present the reproduced DMFT equation as a novel derivation that secretly depends on the paper's own conclusions. The correlation prediction (Eq. 57) is checked against numerical simulation in Fig. 5 as an independent, parameter-free benchmark. The algebraic error in the double-Laplace-transform identity (Eq. 56) flagged by a close reader is a correctness flaw in a tutorial step, not a circular step: it does not define the predicted correlation in terms of itself, and the final C(t-s) result remains externally verifiable against simulation. Therefore no significant circularity is present.

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

No free parameters are fitted; the model parameters sigma, sigma0, mu, and gamma are inputs from the model definitions. The paper introduces the term 'shuffling randomness' as a pedagogical label but no new entity. The main unproved inputs are standard random matrix theory results and the self-averaging assumption in the nonlinear GLV case.

assumptions (4)

standard math The spectral density of the random matrix A is given by the Wigner semicircle law, rotated onto the imaginary axis for the skew-symmetric case.
Used in Section 2.4.1 (Eqs. 14-17) to replace eigenvalue sums by integrals over the spectral density. This is a standard random matrix theory result reviewed in Appendix A.
domain assumption Shuffled averages of order parameters converge to quenched averages in the large-N limit.
Proven for the linear model in Section 2.4.2, but asserted without proof for the nonlinear GLV model in Section 5.2.5 (Eq. 81). The DMFT closure depends on it.
domain assumption The cavity perturbation Ai0 x0(t) is O(1/sqrt(N)) and linear response captures the leading-order effect exactly.
Section 3.2 and Section 5.2.2 (Eq. 73). This is the standard DMFT cavity assumption; the paper provides order-of-magnitude estimates rather than a rigorous uniform bound.
domain assumption Entries Aij are Gaussian, so sums over A0i are Gaussian processes and Wick's theorem applies.
Footnote 2 in Section 2.1 says Gaussianity is not strictly necessary but simplifies analysis. Wick's theorem is used in Sections 3.4 and 5.2.4.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Building Intuition for Dynamical Mean-Field Theory: A Simple Model and the Cavity Method." pith.science (2026). https://pith.science/paper/VLCN6RGF

@misc{pith2026250716654,
  author       = {Pith},
  title        = {Pith review of: Building Intuition for Dynamical Mean-Field Theory: A Simple Model and the Cavity Method},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VLCN6RGF}},
  note         = {Machine review of arXiv:2507.16654}
}

read the original abstract

Dynamical Mean-Field Theory (DMFT) is a powerful theoretical framework for analyzing systems with many interacting degrees of freedom. This tutorial provides an accessible introduction to DMFT. We begin with a linear model where the DMFT equations can be derived exactly, allowing readers to develop clear intuition for the underlying principles. We then introduce the cavity method, a versatile approach for deriving DMFT equations for non-linear systems. The tutorial concludes with an application to the generalized Lotka--Volterra model of interacting species, demonstrating how DMFT reduces the complex dynamics of many-species communities to a tractable single-species stochastic process. Key insights include understanding how quenched disorder enables the reduction from many-body to effective single-particle dynamics, recognizing the role of self-averaging in simplifying complex systems, and seeing how collective interactions give rise to non-Markovian feedback effects.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Towards Chemically Accurate and Scalable Quantum Simulations on IQM Quantum Hardware: A Quantum-HPC Hybrid Approach
quant-ph 2026-04 accept novelty 5.5 of 10

SQD with LUCJ (and a new LCNot-UCCSD variant) on IQM Sirius recovers chemically accurate energies, 1D/2D PES, and DMET-embedded ligand/amantadine results versus FCI/CASCI references.
Machine-Learned Compact Subspace Generation for Quantum Selected Configuration Interaction within Density Matrix Embedding Framework
quant-ph 2026-07 conditional novelty 4.0 of 10

An RBM-guided selected-CI solver inside DMET reaches the DMET-CASCI energy within 1.6 mHa using ~4% of the symmetry-valid configuration subspace on an 11-fragment protein–ligand model.

Reference graph

Works this paper leans on

14 extracted references · 8 canonical work pages · cited by 2 Pith papers

[3]

& Cammarota, C

Roy, F., Biroli, G., Bunin, G. & Cammarota, C. Numerical implementation of dynamical mean field theory for disordered systems: application to the lotka–volterra model of ecosystems.Journal of Physics A: Mathematical and Theoretical52, 484001 (2019). URLhttps://dx.doi.org/10.1088/ 1751-8121/ab1f32

work page 2019
[1]

Statistical Physics of Spin Glasses and Information Processing: An Introduction (Oxford University Press, 2001)

Nishimori, H. Statistical Physics of Spin Glasses and Information Processing: An Introduction (Oxford University Press, 2001). URL https://doi.org/10.1093/acprof:oso/9780198509417.001.0001. https://academic.oup.com/book/5185/book-pdf/54038185/ acprof-9780198509400.pdf

arXiv 2001
[2]

Cui, W., III, R. M. & Mehta, P. Les houches lectures on community ecology: From niche theory to statistical mechanics (2024). URLhttps: //arxiv.org/abs/2403.05497. 2403.05497

arXiv 2024
[4]

Cui, W., Rocks, J. W. & Mehta, P. An elementary mean-field approach to the spectral densities of random matrix ensembles.Physica A: Statistical Mechanics and its Applications637, 129608 (2024). URLhttps://www. sciencedirect.com/science/article/pii/S037843712400116X

work page 2024
[5]

Ecological communities with lotka-volterra dynamics.Phys

Bunin, G. Ecological communities with lotka-volterra dynamics.Phys. Rev. E95, 042414 (2017). URLhttps://link.aps.org/doi/10.1103/ PhysRevE.95.042414

work page 2017
[6]

& Mehta, P

Cui, W., Marsland, R. & Mehta, P. Effect of resource dynamics on species packing in diverse ecosystems.Phys. Rev. Lett.125, 048101 (2020). URL https://link.aps.org/doi/10.1103/PhysRevLett.125.048101

work page doi:10.1103/physrevlett.125.048101 2020
[7]

& Bouchaud, J.A First Course in Random Matrix Theory: for Physicists, Engineers and Data Scientists(Cambridge University Press, 2020)

Potters, M. & Bouchaud, J.A First Course in Random Matrix Theory: for Physicists, Engineers and Data Scientists(Cambridge University Press, 2020). URL https://books.google.com/books?id=9K-WzQEACAAJ

work page 2020
[8]

& Huang, H

Zou, W. & Huang, H. Introduction to dynamical mean-field theory of randomly connected neural networks with bidirectionally correlated couplings. SciPost Phys. Lect. Notes79 (2024). URL https://scipost. org/10.21468/SciPostPhysLectNotes.79. Dynamical Mean-Field Theory Tutorial 33

work page doi:10.21468/scipostphyslectnotes.79 2024

Show all 14 references

[9]

& Sompolinsky, H

Crisanti, A. & Sompolinsky, H. Path integral approach to random neural networks. Phys. Rev. E 98, 062120 (2018). URL https://link.aps. org/doi/10.1103/PhysRevE.98.062120

2018 doi
[10]

& Sommers, H

Sompolinsky, H., Crisanti, A. & Sommers, H. J. Chaos in random neural networks. Phys. Rev. Lett.61, 259–262 (1988). URLhttps://link.aps. org/doi/10.1103/PhysRevLett.61.259

1988 doi
[11]

& Sompolinsky, H

Kadmon, J. & Sompolinsky, H. Transition to chaos in random neuronal networks. Phys. Rev. X5, 041030 (2015). URLhttps://link.aps.org/ doi/10.1103/PhysRevX.5.041030

2015 doi
[12]

& Helias, M

Schuecker, J., Goedeke, S. & Helias, M. Optimal sequence memory in driven random networks.Phys. Rev. X8, 041029 (2018). URLhttps: //link.aps.org/doi/10.1103/PhysRevX.8.041029

2018 doi
[13]

& Bunin, G

Arnoulx de Pirey, T. & Bunin, G. Many-species ecological fluctuations as a jump process from the brink of extinction.Phys. Rev. X14, 011037 (2024). URL https://link.aps.org/doi/10.1103/PhysRevX.14.011037

2024 doi
[14]

Novikov, E. A. Functionals and the random-force method in turbulence theory. Sov. Phys. JETP20, 1290–1294 (1965)

1965

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [3]

& Cammarota, C

Roy, F., Biroli, G., Bunin, G. & Cammarota, C. Numerical implementation of dynamical mean field theory for disordered systems: application to the lotka–volterra model of ecosystems.Journal of Physics A: Mathematical and Theoretical52, 484001 (2019). URLhttps://dx.doi.org/10.1088/ 1751-8121/ab1f32

work page 2019

[2] [1]

Statistical Physics of Spin Glasses and Information Processing: An Introduction (Oxford University Press, 2001)

Nishimori, H. Statistical Physics of Spin Glasses and Information Processing: An Introduction (Oxford University Press, 2001). URL https://doi.org/10.1093/acprof:oso/9780198509417.001.0001. https://academic.oup.com/book/5185/book-pdf/54038185/ acprof-9780198509400.pdf

arXiv 2001

[3] [2]

Cui, W., III, R. M. & Mehta, P. Les houches lectures on community ecology: From niche theory to statistical mechanics (2024). URLhttps: //arxiv.org/abs/2403.05497. 2403.05497

arXiv 2024

[4] [4]

Cui, W., Rocks, J. W. & Mehta, P. An elementary mean-field approach to the spectral densities of random matrix ensembles.Physica A: Statistical Mechanics and its Applications637, 129608 (2024). URLhttps://www. sciencedirect.com/science/article/pii/S037843712400116X

work page 2024

[5] [5]

Ecological communities with lotka-volterra dynamics.Phys

Bunin, G. Ecological communities with lotka-volterra dynamics.Phys. Rev. E95, 042414 (2017). URLhttps://link.aps.org/doi/10.1103/ PhysRevE.95.042414

work page 2017

[6] [6]

& Mehta, P

Cui, W., Marsland, R. & Mehta, P. Effect of resource dynamics on species packing in diverse ecosystems.Phys. Rev. Lett.125, 048101 (2020). URL https://link.aps.org/doi/10.1103/PhysRevLett.125.048101

work page doi:10.1103/physrevlett.125.048101 2020

[7] [7]

& Bouchaud, J.A First Course in Random Matrix Theory: for Physicists, Engineers and Data Scientists(Cambridge University Press, 2020)

Potters, M. & Bouchaud, J.A First Course in Random Matrix Theory: for Physicists, Engineers and Data Scientists(Cambridge University Press, 2020). URL https://books.google.com/books?id=9K-WzQEACAAJ

work page 2020

[8] [8]

& Huang, H

Zou, W. & Huang, H. Introduction to dynamical mean-field theory of randomly connected neural networks with bidirectionally correlated couplings. SciPost Phys. Lect. Notes79 (2024). URL https://scipost. org/10.21468/SciPostPhysLectNotes.79. Dynamical Mean-Field Theory Tutorial 33

work page doi:10.21468/scipostphyslectnotes.79 2024