Pith. sign in

REVIEW 4 major objections 5 minor 37 references

Statistical inference for mean-field queueing systems

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A closed-form least-squares estimator recovers the arrival and service rates of a large power-of-two queueing network, with proven consistency and asymptotic normality.

desk verdict The paper's LSE is built on a service drift that leaks probability mass, so its main theorems do not describe the supermarket model. read the letter →

arxiv 2411.12936 v1 pith:6B4L3BKO submitted 2024-11-20 math.ST math.PRstat.TH

Ioannis Lambadaris , Ahmed Sid-Ali , Wei Sun , Yiqiang Q. Zhao This is my paper · ORCID
classification math.STmath.PRstat.TH MSC 60K2560K3562F1062F12
keywords mean-fieldqueueingsupermarketmodelpoweroftwochoicesleastsquaresestimationconsistencyasymptoticnormalityempiricalmeasureparameter
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

The paper asks whether the arrival rate $\lambda$ and service rate $\nu$ of a large supermarket-style queueing network can be estimated from discrete-time snapshots of the aggregate empirical measure, without observing individual servers. It constructs an approximate least-squares estimator from the mean-field ODE limit and proves that the estimator is consistent and asymptotically normal as both the number of servers $N$ and the number of observation times $m$ grow. This is presented as the first statistical inference scheme for a discrete-state mean-field queueing model, where the fluctuations are governed by a Gaussian process rather than by Brownian diffusion. If the main theorems are right, operators of large load-balancing networks can recover the fundamental arrival and service rates from coarse aggregate observations.

What carries the argument

The machinery is the linear-in-parameter decomposition of the mean-field drift, $F(x)=\lambda\sum_j U_j(x)e_j+\nu\sum_j V_j(x)e_j$, with $U_j$ and $V_j$ given by (3.1)-(3.2). Because $\lambda$ and $\nu$ enter linearly, the minimizer of the quadratic criterion $G(\lambda,\nu)$ has the closed form (3.6) in terms of integrals of $U_j$ and $V_j$ along the limiting trajectory; replacing those integrals by Riemann sums over the observed empirical measure produces the estimator (3.10). The argument also rests on the law of large numbers for the empirical measure (Theorem 2.1), the central limit theorem whose limiting process $Z$ solves the SDE (2.5), and the identifiability condition $a_{11}a_{22}>(a_{12})^2$, which keeps the $2\times2$ normal matrix invertible.

What would settle it

Take any true parameter pair $(\lambda^*,\nu^*)$ for the power-of-two model, solve the mean-field ODE (2.1) numerically on $[0,T]$, compute $a_{11},a_{12},a_{22}$, and verify the strict inequality $a_{11}a_{22}>(a_{12})^2$; then simulate the $N$-server process with large $N$ and $m$ satisfying $m/\sqrt{N}\to\infty$. If for any such parameter pair the estimates fail to concentrate at $(\lambda^*,\nu^*)$, or the $\sqrt{N}$-rescaled errors fail to match the Gaussian law of Theorem 3.2, the central claim falls.

Watch

Extended reading notes

Core claim

The central claim is that, for the power-of-two supermarket model, the unknown rates $\theta^*=(\lambda^*,\nu^*)$ are identifiable from the finite dataset $D_{N,m}=\{\varrho^N(t_k):1\le k\le m\}$ and can be recovered by the closed-form least-squares formula (3.10). Theorem 3.1 states that, under the strict identifiability condition $a_{11}a_{22}>(a_{12})^2$, the estimator converges in probability to $\theta^*$ as $N,m\to\infty$. Theorem 3.2 states that $\sqrt{N}(\hat\theta-\theta^*)$ converges in distribution to a Gaussian vector as $N,m\to\infty$ with $m/\sqrt{N}\to\infty$, and the limiting law is a linear combination of the Gaussian fluctuation process $Z$ arising from the central limit theorem for the empirical measure. The proof structure carries all randomness through $Z$, so the estimator's asymptotic distribution inherits Gaussianity from Proposition 2.1.

Load-bearing premise

Everything rests on the strict identifiability condition $a_{11}a_{22}>(a_{12})^2$, which is stated in terms of the unknown true mean-field trajectory; if it fails, the normal matrix in the estimator is singular and the whole scheme is undefined.

Editorial extensions

If this is right

  • The estimator (3.10) is a closed-form function of aggregate queue-length snapshots, so it can be applied when tracking every server individually is impractical.
  • As both the system size and the number of observation epochs grow, the estimates concentrate at the true arrival and service rates, as stated in Theorem 3.1.
  • The $\sqrt{N}$-scaled estimation error converges to a Gaussian distribution, so the authors' expressions for its mean and covariance can in principle yield confidence intervals for $\lambda$ and $\nu$.
  • The authors note that the same approximate-LSE scheme is conceivable for mean-field models whose limiting equation is linear in the unknown parameters, such as load-balancing cloud storage models, provided the limiting equation remains tractable.
  • The linear-in-parameter structure is essential: for nonparametric estimation of interaction kernels the optimization moves to function space, and the paper's approach does not directly carry over.

Reading between the lines

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

  • Beyond the paper, the Gaussian limit suggests a practical monitoring device: sliding-window estimates of $\lambda$ and $\nu$ from aggregate queue-length histograms could detect changes in load or service capacity, with standard errors read off the limiting covariance.
  • Because the identifiability condition is stated in terms of the unknown mean-field trajectory, a data-driven implementation would likely need to check the sample Gram matrix for near-singularity; poor conditioning would show up as unstable estimates rather than honest uncertainty.
  • The requirement $m/\sqrt{N}\to\infty$ means the observation grid must be fine relative to the square root of the network size, which quantifies how much discretization bias the fluctuation analysis tolerates.
  • The authors' numerical results show skewness and kurtosis close to Gaussian values at moderate $N$ and $m$, suggesting the asymptotic regime may be reached earlier in practice than the theorem's joint limit requires.
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 / 5 minor

Summary. The paper proposes an approximate least-squares estimator (LSE) for the arrival rate λ and service rate ν of the power-of-two supermarket model, based on discrete-time observations of the empirical measure of queue lengths over a fixed interval [0,T]. The estimator is obtained by matching the observed increments of the empirical measure to the mean-field ODE recalled in Theorem 2.1. The authors claim consistency (Theorem 3.1) and asymptotic normality (Theorem 3.2) as the number of servers N and the number of observations m tend to infinity, using the existing law of large numbers and central limit theorem for the supermarket model. Numerical experiments with simulated data are reported to support the theoretical results.

Significance. If correct, the paper would address a genuinely open problem: statistical inference for discrete-state mean-field queueing models, as opposed to the more developed inference theory for interacting diffusions. The construction is natural—regress empirical increments on the mean-field vector field—and the use of the external LLN/CLT results is appropriate and not circular. The paper also provides explicit proofs and numerical validation, which is a strength. However, the central estimator is built from a vector field that does not preserve total mass, and the consistency/normality proofs contain convergence arguments that are not justified by the cited results. The claimed theorems therefore do not hold for the estimator as defined in Section 3.2, and the paper needs substantial correction before its central claim can be accepted.

major comments (4)
  1. [Section 2.2, Eq. (2.7); Section 3.2, Eq. (3.2)] The service term in the mean-field vector field is incorrect at coordinate 0 and breaks mass conservation. For any probability vector x, V_j(x)=x_{j+1}-x_j gives Σ_{j=0}^∞ V_j(x) = -x_0, so Σ_{j=0}^∞ F_j(x) = -ν x_0 ≠ 0 when x_0>0. A service completion at a queue of length 1 increases the mass at state 0 by ν x_1, while no service completion can occur at length 0; hence the correct generator must have V_0(x)=x_1, not x_1-x_0. With that correction, Σ_j V_j(x)=0. As written, the ODE (2.1) cannot remain in the simplex S, contradicting Theorem 2.1. Because the LSE (3.10), the normal equations (3.4)-(3.5), and the definitions of a_12, a_22, b_1, b_2 all use the same erroneous V_j, the estimator targets the wrong mean-field limit. Consequently Theorem 3.1 cannot hold for the estimator as defined, and Theorem 3.2 inherits the same problem.
  2. [Appendix C, Eq. (C.1)] The proof of Theorem 3.1 asserts that sup_{t∈[0,T]} Σ_{j=0}^∞ |ϱ^N_j(t)-ϱ_j(t)| → 0 in probability. This is a uniform-in-time l1 convergence statement, but Theorem 2.1 only provides convergence in distribution in D([0,T],S) in the Skorokhod topology. That topology does not by itself yield the uniform l1 bound used in the subsequent estimates for a^N,m_11, b^N,m_1, etc. A uniform tail estimate or an explicit statement of the metric on S and a continuity argument for the limit is needed; without it, the consistency proof has a gap.
  3. [Appendix D, Eq. (D.5)] The proof of Theorem 3.2 passes from convergence of Z^N in D([0,T],ℓ2) to the uniform convergence statement sup_t ∥Z^N(t)-Z(t)∥_2 → 0 in probability, and then uses quantities such as sup_t Σ_i |Z^N_i(t)| together with the weighted ℓ2 moment bound (D.2). Skorokhod convergence does not automatically imply the displayed uniform sup-norm convergence, and the sup of the l1 norm is not controlled by the second-moment bound used. These steps need a rigorous justification or a different argument for the convergence of the terms in (D.1).
  4. [Section 3.2, Lemma 3.1] The identifiability condition a_11 a_22 > (a_12)^2 in (3.7) is stated in terms of the unknown mean-field solution ϱ(·,θ*), and the sufficient conditions (3.8) and (3.9) also involve integrals of the unknown ϱ. If (3.7) fails, the matrix in (3.10) is singular and the estimator is undefined. The paper should discuss how this condition can be checked or avoided in practice; currently it is an assumption that cannot be verified from the data before estimation.
minor comments (5)
  1. [Abstract] The word 'mean-filed' in the abstract should be 'mean-field'.
  2. [Theorem 3.2] The statement 'as N, m, m√N → ∞' is ambiguous; based on the proof in Appendix D, the intended condition is m/√N → ∞, and the text should be corrected.
  3. [Table 1] In the row for (N,m)=(1000,10000), the first entry is written as '0.28, 0.74)' with a missing opening parenthesis.
  4. [References] Reference [15] is cited for Ethier-Kurtz results used in (C.1) and (D.5), but the listed title 'The infinitely-many-alleles model with selection as a measure-valued diffusion' is not the standard Ethier-Kurtz text; the citation needs to be corrected and the page references checked.
  5. [Section 3.2] The sums in (3.10) run over all j∈Z+, but any numerical implementation must truncate the state space; the paper does not discuss the effect of truncation on the estimator, which would be relevant for the simulation results in Section 4.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the LSE and its asymptotics are derived from externally cited LLN/CLT results via a standard delta method, and the only self-citation is motivational.

full rationale

The derivation chain is not circular. The mean-field ODE (2.1)-(2.7) and the fluctuation SDE (2.5) are imported from Graham [20, Theorem 3.4] and Budhiraja-Friedlander [6, Theorem 2 and Proposition 2], neither of which overlaps with the present authors; these are external independent results, so Theorem 2.1 and Theorem 2.2 are real evidence rather than self-citation. Proposition 2.1 is proved in Appendix A from that SDE and from estimates in [6], so the Gaussianity used in Theorem 3.2 is derived, not assumed. The approximate LSE (3.10) is a plug-in version of the normal equations (3.6) obtained by minimizing G(λ,ν); its consistency and asymptotic normality follow by standard delta-method arguments in Appendices C-D from the same external LLN/CLT. The identifiability condition (3.7) is a regularity assumption about the true limiting trajectory, not a hidden use of the estimator's output, and the sufficient conditions in Lemma 3.1 are proved directly. The only self-citation, [37], appears in Section 1 solely to assert that statistical inference for mean-field queueing models is largely unexplored; it is motivational and does not support the theorems. The possible mass-conservation defect in F noted by a skeptic is a correctness concern about the recalled ODE, not a circular reduction of the estimator to its own inputs, and therefore falls outside the circularity scoring under the hard rules.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

The central claim rests on the established mean-field LLN and CLT for the supermarket model, on an identifiability condition that is not verifiable from data, and on the assumption of full empirical measure observations. No ad hoc free parameters or invented entities are introduced; the unknown arrival and service rates are the legitimate targets of inference.

assumptions (7)
  • domain assumption The mean-field law of large numbers, Theorem 2.1 from Graham, holds: rho^N converges to rho in probability in D([0,T],S), where rho solves the ODE (2.1).
    Invoked in the consistency proof in Section 3.3 and in the proof of Theorem 3.2; not proved in this paper.
  • domain assumption The mean-field central limit theorem, Theorem 2.2 from Budhiraja and Friedlander, holds: Z^N = sqrt(N)(rho^N - rho) converges in distribution in D([0,T],l2) to the solution of the SDE (2.5), under the stated second-moment conditions.
    Used in the proof of Theorem 3.2 in Appendix D to establish the limiting distribution of the estimator.
  • domain assumption The ODE (2.1) has a unique solution in C([0,T],S) for each initial condition, from Graham's Theorem 3.4.
    Ensures the limiting trajectory used by the estimator exists and is continuous.
  • domain assumption The identifiability condition a11 a22 > (a12)^2 holds; Lemma 3.1 gives sufficient conditions (3.8) and (3.9).
    Both theorems assume (3.7); the sufficient conditions involve the unknown limit rho, so they cannot be verified a priori.
  • domain assumption The data consist of the full empirical measure rho^N(t_k) for all queue lengths j in Z+ at each discrete time; there is no measurement error or missing state information.
    The estimator (3.10) sums over all j in Z+ and requires the complete distribution.
  • standard math Standard probabilistic tools hold: Skorokhod representation, continuous mapping theorem, Fatou's lemma, Cauchy-Schwarz inequality, and the equivalence of weak and l1 convergence for probability measures on Z+ as in Lemma 3.2.
    Used throughout the proofs in Appendices A through D.
  • domain assumption Initial conditions satisfy the hypotheses of Theorem 2.2: sup_N sum_j j^2 rho^N_j(0) is finite, Z^N(0) converges to z0 in l2, and sup_N sum_j j^2 (Z^N_j(0))^2 is finite.
    These are stated in Theorem 2.2 and inherited by Theorem 3.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Statistical inference for mean-field queueing systems." pith.science (2026). https://pith.science/paper/6B4L3BKO

@misc{pith2026241112936,
  author       = {Pith},
  title        = {Pith review of: Statistical inference for mean-field queueing systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6B4L3BKO}},
  note         = {Machine review of arXiv:2411.12936}
}
read the original abstract

Mean-field limits have been used now as a standard tool in approximations, including for networks with a large number of nodes. Statistical inference on mean-filed models has attracted more attention recently mainly due to the rapid emergence of data-driven systems. However, studies reported in the literature have been mainly limited to continuous models. In this paper, we initiate a study of statistical inference on discrete mean-field models (or jump processes) in terms of a well-known and extensively studied model, known as the power-of-L, or the supermarket model, to demonstrate how to deal with new challenges in discrete models. We focus on system parameter estimation based on the observations of system states at discrete time epochs over a finite period. We show that by harnessing the weak convergence results developed for the supermarket model in the literature, an asymptotic inference scheme based on an approximate least squares estimation can be obtained from the mean-field limiting equation. Also, by leveraging the law of large numbers alongside the central limit theorem, the consistency of the estimator and its asymptotic normality can be established when the number of servers and the number of observations go to infinity. Moreover, numerical results for the power-of-two model are provided to show the efficiency and accuracy of the proposed estimator.

Figures

Figures reproduced from arXiv: 2411.12936 by the authors.

Figure 1
Figure 1. Values of 100 Least squares estimates of the parameter [PITH_FULL_IMAGE:figures/full_fig_p013_1.png] view at source ↗
Figure 2
Figure 2. Values of 100 Least squares estimates of the parameter [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Histograms of the normalized errors √ N(λ N,m − λ) with the associated kernel density esti￾mate plots for different values of the network size N and the number of observations m with the true parameters θ = (λ, ν) = (0.5, 1) 17 [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Histograms of the normalized errors √ N(ν N,m −ν) with the associated kernel density estimate plots for different values of the network size N and the number of observations m, corresponding to true parameters θ = (λ, ν) = (0.5, 1) 5 Conclusions and perspectives In thi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages

  1. [1]

    Amorino, A

    C. Amorino, A. Heidari, V. Pilipauskaite, and M. Podolskij. Parameter estimation of discretely observed interacting particle systems. Stochastic Processes and their Applications , 163:350–386, 2023

    work page 2023

  2. [2]

    Asanjarani, Y

    A. Asanjarani, Y. Nazarathy, and P. Taylor. A survey of parameter and state estimation in queues. Queueing Syst , 97:39–80, 2021

    work page 2021

  3. [3]

    Belomestny, V

    D. Belomestny, V. Pilipauskaite, and M. Podolskij. Semiparametric estimation of mckean-vlasov sdes. Annales de l’Institut Henri Poincar´ e, Probabilit´ es et Statistiques, 59(1):79–96, 2023

    work page 2023

  4. [4]

    Bena ¨ ım and J.Y

    M. Bena ¨ ım and J.Y. Le Boudec. A class of mean field interaction models for computer and communication systems. Performance Evaluation, 65(11–12):823–838, 2008

    work page 2008

  5. [5]

    P.N. Bishwal. Estimation in interacting diffusions: Continuous and discrete sampling. Applied Mathematics-a Journal of Chinese Universities Series B , 02:1154–1158, 2011

    work page 2011

  6. [6]

    Budhiraja and E

    A. Budhiraja and E. Friedlander. Diffusion approximations for load balancing mechanisms in cloud storage systems. Advances in Applied Probability, 51(1):41–86, 2019

    work page 2019

  7. [7]

    Budhiraja, D

    A. Budhiraja, D. Mukherjee, and R. Wu. Supermarket model on graphs. The Annals of Applied Probability, 29(3):1740–1777, 2019

    work page 2019

  8. [8]

    P. Curie. Magnetic properties of materials at various temperatures. Ann. Chem. Phys , 5(289), 1895

    work page

Show all 37 references
  1. [9]

    D.A. Dawson. Critical dynamics and fluctuations for a mean field model of cooperative behaviour. J. Statist. Phys. , 41:29–85, 1983

    1983

  2. [10]

    D.A. Dawson. Introductory lectures on stochastic population systems. arXiv:1705.03781 [math.PR], 2017

    2017 arXiv

  3. [11]

    Dawson, J

    D.A. Dawson, J. Tang, and Y.Q. Zhao. Balancing queues by mean field interactions. Queueing Syst, 49:335–361, 2005

    2005

  4. [12]

    Dawson, J

    D.A. Dawson, J. Tang, and Y.Q. Zhao. Performance analysis of joining the shortest queue model among a large number of queues. Asia-Pacific Journal of Operational Research , 36(4), 2019

    2019

  5. [13]

    Delattre and N

    S. Delattre and N. Fournier. Statistical inference versus mean field limit for hawkes processes. Electronic Journal of Statistics , 10(1):1223–1295, 2016

    2016

  6. [14]

    Dobrushin and Y.M

    R.L. Dobrushin and Y.M. Sukhov. Asymptotic investigation of star-shaped message switching networks with a large number of radial rays. Probl. Inf. Trans., 12(1):49–66, 1976

    1976

  7. [15]

    Ethier and T.G

    S.N. Ethier and T.G. Kurtz. The infinitely-many-alleles model with selection as a measure-valued diffusion. Stochastic Methods in Biology, volume 70. Springer-Verlag, Berlin-Heidelberg-N.Y., 1987

    1987

  8. [16]

    G¨ artner

    J. G¨ artner. On the mckean-vlasov limit for interacting diffusions.Math. Nachr., 137:197–248, 1988. 30

    1988

  9. [17]

    Genon-Catalot and C

    V. Genon-Catalot and C. Lar´ edo. Parametric inference for small variance and long time horizon mckean-vlasov diffusion models. Electronic Journal of Statistics , 15(2):5811–5854

  10. [18]

    Giesecke, G

    K. Giesecke, G. Schwenkler, and J.A. Sirignano. Inference for large financial systems. Math. Fin. , 30:823–838, 2019

    2019

  11. [19]

    Giesecke, K

    K. Giesecke, K. Spiliopoulos, R.B. Sowers, and J.A. Sirignano. Large portfolio asymptotics for loss from default. Mathematical Finance, 25(1):77–114, 2015

    2015

  12. [20]

    C. Graham. Chaoticity on path space for a queueing network with selection of the shortest queue amongst several. J. Appl. Prob. , 37(1):198–211, 2000

    2000

  13. [21]

    Graham and S

    C. Graham and S. M´ el´ eard. Propagation of chaos for a fully connected loss network with alternate routing. Stoch. Processes Appl., 44:159–180, 1993

    1993

  14. [22]

    Graham and S

    C. Graham and S. M´ el´ eard. Dynamic asymptotic results for a generalized star-shaped ioss network. Ann. Appl. Prob. , 5, 1995

    1995

  15. [23]

    McKean Jr

    H.P. McKean Jr. A class of markov processes associated with nonlinear parabolic equations. Proc. Natl. Acad. Sci. USA , 56(6):1907–1911, 1966

    1907

  16. [24]

    McKean Jr

    H.P. McKean Jr. Speed of approach to equilibrium for kac’s caricature of a maxwellian gas. Arch. Ration. Mech. Anal., 21(5):343–367, 1966

    1966

  17. [25]

    M. Kac. Foundations of kinetic theory. In Calif University of California Press, Berkeley, editor, Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, Volume 3: Contributions to Astronomy and Physics , pages 171–197, 1956

    1956

  18. [26]

    R.A. Kasonga. Maximum likelihood theory for large interacting systems. SIAM Journal on Applied Mathematics, 50(3):865–875, 1990

    1990

  19. [27]

    O. Kley, C. Kl¨ uppelberg, and L. Reichel. Systemic risk through contagion in a core-periphery structured banking network. Banach Center publications , 104:133–149, 2015

    2015

  20. [28]

    C. Liu. Statistical inference for a partially observed interacting system of hawkes processes.Stochas- tic Processes and their Applications , 130(9):5636–5694, 2020

    2020

  21. [29]

    Della Maestra and M

    L. Della Maestra and M. Hoffmann. Nonparametric estimation for interacting particle systems: Mckean-vlasov models. Probab. Theory Relat. Fields , 182:551–613, 2022

    2022

  22. [30]

    M´ el´ eard and V

    S. M´ el´ eard and V. Bansaye.Some Stochastic Models for Structured Populations: Scaling Limits and Long Time Behavior . Springer, 2015

    2015

  23. [31]

    Mitzenmacher

    M. Mitzenmacher. The power of two choices in randomized load balancing. PhD thesis, University of California at Berkeley , 1996

    1996

  24. [32]

    Sharrock, N

    L. Sharrock, N. Kantas, P. Parpas, and G.A. Pavliotis. Online parameter estimation for the mckean- vlasov stochastic differential equation. Stochastic Processes and their Applications , 162:481–546, 2023. 31

    2023

  25. [33]

    J.M. Steele. The Cauchy-Schwarz Master Class . Cambridge University Press, 2004

    2004

  26. [34]

    A. W. van der Vaart. Asymptotic Statistics . Cambridge Series in Statistical and Probabilistic Mathematics. Cambridge University Press, 1998

    1998

  27. [35]

    Vvedenskaya, R.L

    N.D. Vvedenskaya, R.L. Dobrushin, and F.I. Karpelevich. Queueing system with selection of the shortest of two queues: An asymptotic approach. Problems Inform. Transmission , 32(1):15–27, 1996

    1996

  28. [36]

    P. Weiss. L’hypothese du champ mol´ eculaire et la propri´ et´ e ferromagnetique. J. Phys. Theor. Appl., 6(1):661–690, 1907

    1907

  29. [37]

    Y.Q. Zhao. Statistical inference for mean-field queueing models. Queueing Syst, 100:569–571, 2022. 32

    2022

Pith tools

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