Pith. sign in

REVIEW 2 major objections 5 minor 28 references

Neural variational solvers recover known Young-diagram limit shapes and give numerical evidence for a deformation-dependent saddle family when no analytical profile is assumed.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-30 12:19 UTC pith:2CN5C5XV

load-bearing objection Careful methods paper: ensemble-adapted neural variational solvers plus honest three-way numerics on a deformed hook ensemble, not a closed-form limit-shape theorem. the 2 major comments →

arxiv 2607.27061 v1 pith:2CN5C5XV submitted 2026-07-29 cond-mat.stat-mech math-phmath.MPphysics.data-an

Neural variational framework for random Young-diagram limit shapes

Qian Chen , Bo-Xuan Ge This is my paper
classification cond-mat.stat-mech math-phmath.MPphysics.data-an MSC 05E1060C0582B05 PACS 05.10.-a05.20.-y02.70.Rr
keywords Young diagramslimit shapesPlancherel measureneural variational methodshook-length formulaq-PlancherelMetropolis-Hastingsexclusion statistics
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Random Young diagrams drawn from different probability measures often concentrate, at large size, on a deterministic macroscopic profile. This paper builds a structure-preserving neural variational method that chooses its variables to match each ensemble: soft hook actions for hook-weighted measures, entropy densities for uniform and exclusion ensembles, and discrete row fractions when the leading rows scale linearly with n. The method is checked on Plancherel, uniform, minimal-difference, and fixed-q q-Plancherel ensembles, recovering the known shapes without feeding those shapes into training. It is then applied to a quartically deformed hook-length ensemble that has no assumed closed-form saddle. Large-n neural profiles, finite-n exact-action maximum-probability diagrams, and finite-temperature Metropolis–Hastings mean profiles all show the same trend—stronger deformation shortens the longest rows and spreads boxes over more rows—and agree at the percent level. The agreement supplies numerical evidence for a deformation-dependent family of macroscopic saddles.

Core claim

For a quartically deformed hook-length ensemble with no assumed analytical saddle, large-n neural profiles, finite-n exact-action MAP profiles, and T=1 corner-transfer MCMC mean profiles agree at the percent level and share the ordered deformation trend that increasing the quartic strength suppresses the leading rows and broadens the support, giving numerical evidence for a common macroscopic saddle at each deformation strength.

What carries the argument

A structure-preserving neural variational framework: ensemble-adapted parameterizations (continuous monotone rows with soft hook occupancy, density-first Bose/exclusion entropy, or discrete row fractions) that enforce nonnegativity, monotonicity, and fixed area, optimized on the defining finite-size action or continuum entropy and checked after the fact against exact integer actions, MAP search, and sampling.

Load-bearing premise

Percent-level agreement among a continuous neural relaxation at huge n, integer maximum-probability diagrams only up to a few thousand boxes, and finite-temperature samples at those same modest sizes is taken as enough evidence that a unique macroscopic limiting saddle exists for each deformation strength.

What would settle it

Push the discrete MAP search and corner-transfer MCMC to substantially larger n for several positive deformation strengths; if the mean and MAP profiles peel away from each other or from the large-n neural profile, or if a second macroscopic shape appears with comparable action, the claimed common saddle family fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Known Plancherel, uniform, minimal-difference, and fixed-q q-Plancherel limit shapes can be recovered from the defining action or entropy without using the analytical profiles in training.
  • For the quartic hook ensemble, stronger deformation systematically shortens the longest rows and spreads mass over more rows.
  • MAP diagrams and typical sampled means can sit in the same macroscopic saddle region even when they are not identical by definition.
  • The same structure-preserving strategy can be reused for other deformed or nonlocal Young-diagram measures that lack closed-form saddles.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A continuum Euler–Lagrange equation for the quartic deformation, once derived, should be directly testable against the reported neural profiles.
  • Competing or metastable saddles under stronger or multi-well deformations would be a natural next stress test of whether the method can detect non-uniqueness.
  • The ch=1 normalization makes the undeformed MAP coincide with Plancherel while changing the finite-temperature measure; repeating the triad of calculations at the strict Plancherel hook coefficient would tighten the physical interpretation.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper introduces a structure-preserving neural variational framework for random Young-diagram ensembles, with representations chosen to match each measure’s structure and scaling (soft hook actions for Plancherel-type models, density/entropy formulations for uniform and minimal-difference partitions, and discrete row fractions for fixed-q q-Plancherel). Known analytical or asymptotic profiles are used only for post-training validation. The main non-benchmark application is a quartically deformed hook-length ensemble (ch=1) with no assumed saddle: large-n neural profiles (n=2×10^7) are compared with exact-action MAP candidates (n≤8000) and T=1 corner-transfer MCMC means. Increasing θ suppresses leading rows and broadens support, and the three independently obtained profiles agree at about the percent level, offered as numerical evidence for a deformation-dependent macroscopic saddle family.

Significance. If the three-way agreement is read at the strength the authors claim—numerical evidence, not a proof—the work is a solid methodological contribution to asymptotic Young-diagram problems where nonlocal hook actions lack closed saddles. Strengths include: (i) barring analytical profiles, MAP partitions, and MCMC data from training and checkpoint selection; (ii) ensemble-adapted constraints (monotonicity, area, integer projection); (iii) independent cross-checks via exact integer search and equilibrium sampling; and (iv) successful recovery of VKLS, Bose/exclusion entropy shapes, and geometric q-Plancherel row fractions with reported L2 errors from ~10^{-2} to ~10^{-4}. The deformed-model study is a useful template for structure-preserving neural variational methods in combinatorial statistical mechanics.

major comments (2)
  1. [Sec. V.A–V.C, Eq. (80), Tables I–III] Sec. V.A–V.C and Eq. (80): The central deformed-ensemble claim rests on percent-level neural–MAP–MCMC agreement. At θ=0 the MAP sequence still has relative L2 discrepancy 0.147 from VKLS at n=8000 (Sec. V.A), while ENN/MAP_L2=0.0148 uses an area-preserving coarse-graining (Δx=0.05) of the MAP staircase on [0,3.5]. For θ>0 there is no external continuum anchor. Table I shows MAP edge observables stable to ~2% from n=4000 to 8000, and Table III shows MCMC–MAP concentration at the same modest n, but neither establishes that the scaled profile has converged to a unique large-n saddle. Sec. VI already disclaims existence/uniqueness proofs; the abstract and Sec. V conclusions should state more sharply that the evidence is three-way consistency of a deformation trend (including residual finite-size/edge effects and the role of coarse-graining), not demonstrated continuum uniqueness. A short qua
  2. [Sec. II.B.5, Sec. V, Sec. VI] Sec. II.B.5 and Sec. VI: The deformed ensemble uses ch=1 (Gelfand-type hook weight), not the Plancherel normalization ch=2. Undeformed MAP diagrams coincide with Plancherel MAP diagrams, but the finite-T measure sampled by MCMC is not Plancherel. This is stated, yet the abstract and introduction still frame the object as a “quartically deformed hook-length ensemble” adjacent to Plancherel language. For the MCMC–MAP agreement to be interpreted as concentration about the same macroscopic saddle of the studied measure, please keep ch=1 vs ch=2 visually consistent in the abstract/Sec. V headings and avoid any residual implication that the sampled ensemble is a strict Plancherel deformation.
minor comments (5)
  1. [Sec. I] Introduction: “neural-network parameterizations… ans¨ atze” appears to be a UTF-8/encoding glitch for “ansätze”; please fix.
  2. [Figs. 6–9] Fig. 6–9: Coarse-grained MAP curves and fluctuation bands are helpful, but the captions should state explicitly that shaded MCMC bands are four pointwise sample SDs (not SEMs) and that coarse-graining is visualization-only, as already noted in the text.
  3. [Sec. III.B] Eq. (45)–(48): The soft-hook loss omits the overall factor 2 relative to −log P_Pl; this is harmless for minimizers but could be footnoted once next to L_Pl so readers matching to Eq. (15) are not confused.
  4. [Appendix A] Appendix A.2 / Table IV: Principal hyperparameters are well documented; a one-line statement on whether any hyperparameter was tuned using the analytical references (even informally) would further reinforce the post-hoc-only validation claim.
  5. [References] References: Logan–Shepp / Vershik–Kerov classics and the q-Plancherel and exclusion-statistics sources are appropriate; no missing core citations stood out.

Circularity Check

0 steps flagged

No significant circularity: known profiles and discrete MAP/MCMC are barred from training; deformed-saddle claim rests on three separately defined procedures.

full rationale

The paper’s derivation chain does not reduce predictions to their inputs by construction. Benchmark recoveries (Plancherel, uniform, minimal-difference, fixed-q q-Plancherel) minimize ensemble-specific soft-hook or entropy objectives; analytical/asymptotic profiles are used only after training and checkpoint selection (Abstract; Sec. III; Sec. IV). For the quartically deformed ch=1 ensemble, the neural solver at n=2×10^7, the exact-action integer MAP search (n≤8000), and T=1 corner-transfer MCMC are optimized/sampled independently: MAP partitions and MCMC means do not enter the neural loss, initialization, or checkpoints, and no hard-action refinement is applied to projected neural partitions (Sec. III; Sec. V.C; App. A.2.c). Agreement among the three is therefore an empirical cross-check, not a fitted or definitional identity. There is no load-bearing self-citation uniqueness theorem, no parameter fitted to the target shape and re-reported as a prediction, and no renaming of a known deformed profile. Residual concerns about finite-n lag, coarse-graining of MAP staircases, and lack of a proof of existence/uniqueness (Sec. VI) are strength-of-evidence issues, not circularity.

Axiom & Free-Parameter Ledger

7 free parameters · 7 axioms · 2 invented entities

The central numerical claim rests on standard partition/hook combinatorics, large-deviation/variational heuristics for limit shapes, and several modeling and discretization choices internal to the solvers. No new physical entity is postulated. Free parameters are numerical hyperparameters and the deformation/normalization choices that define the studied ensemble family rather than fits to external experimental data.

free parameters (7)
  • quartic deformation strengths θ ∈ {0, 0.1, 0.5, 1} = 0, 0.1, 0.5, 1
    Hand-chosen scan of the new ensemble; the reported saddle family is defined along this grid.
  • hook coefficient ch = 1
    Sets relative weight of the hook term; paper uses ch=1 (Gelfand-type) rather than Plancherel ch=2, which changes the finite-temperature measure.
  • soft-occupancy temperatures τ and stage schedules = e.g. Plancherel 0.8→0.4; quartic 0.8→0.4→0.2→0.1
    Control the continuous relaxation of discrete diagrams; final profiles depend on continuation down to small τ.
  • regularizer weights (wA, wT, wsm, wtail, wC, ...) = ensemble-specific, e.g. quartic wA=1e-1, wT=1e-3
    Weak finite-grid stabilizers in neural objectives; small but nonzero influence on optimized profiles.
  • network width/depth and optimization hyperparameters = 5 hidden layers × 128, seed 1234, ensemble-specific LRs/epochs
    Architecture and AdamW/cosine schedules are design choices that define the variational ansatz capacity.
  • MAP annealing schedule and proposal mixture = T0=2→Tfinal=0.01 over 5e5 steps; mixture probs 0.55/0.18/0.65 etc.
    Heuristic search parameters for approximate integer minimizers; not guaranteed global MAP.
  • MCMC temperature T and sampling budget = T=1; 64 chains × 1e5 production moves, thin 250
    Production uses T=1 with fixed burn-in/thinning; means are finite-sample estimators of the finite-n measure.
axioms (7)
  • standard math Hook-length formula and Plancherel/q-Plancherel normalizations on Young diagrams of Sn.
    Sec. II B; standard representation theory used to define finite-size actions.
  • domain assumption Large-n limit shapes exist as minimizers of ensemble-dependent rate functionals (LDP/variational heuristic).
    Sec. II D writes Pn(f≈f)≍exp(-An Iθ[f]); used as the conceptual target of neural minimization, not proved for the quartic model.
  • domain assumption Bose-type and exclusion-statistics entropy densities select typical profiles for uniform and minimal-difference ensembles.
    Sec. III C citing Vershik and Comtet–Majumdar–Ouvry–Sabhapandit; objectives are built from these continuum entropies.
  • ad hoc to paper Soft occupancy σ((λi−j+δ)/τ) plus softplus hooks is a faithful relaxation of the discrete hook action as τ↓0.
    Sec. III B / App. A 2; central to differentiable training of hook-weighted ensembles.
  • ad hoc to paper Quartic row penalty scaled by √n contributes at the same variational order as the shape-dependent hook term for balanced diagrams.
    Sec. II B 5, Eqs. (27)–(30); justifies the deformed ensemble definition and expected O(1) θ effect.
  • ad hoc to paper Cross-n continuation multi-start row-transfer search finds representative low-action MAP candidates up to n=8000.
    Sec. III E / App. A 5; MAP side of the three-way comparison is only as good as this heuristic global optimization.
  • domain assumption Corner-transfer MH at T=1 mixes sufficiently that thinned means probe the typical saddle region of the finite-n measure.
    Sec. III F / App. A 6; supported by acceptance rates and chain diagnostics but not a mixing proof.
invented entities (2)
  • Quartically deformed hook-length ensemble S(ch)n,θ no independent evidence
    purpose: Provide a non-benchmark hook-weighted measure without assumed analytic saddle for testing the neural framework.
    Defined in Sec. II B 5 with ch and θ; not claimed as a previously studied named ensemble. Independent handle is internal (MAP/MCMC/neural consistency), not an external physical observation.
  • Structure-preserving neural profile parametrizations (tail-integral generator, inverse-EL density, discrete row-fraction nets) independent evidence
    purpose: Enforce Young-diagram constraints and ensemble-appropriate scaling inside the variational ansatz.
    Methodological constructs in Sec. III A–C; validated by recovery of known profiles rather than by external measurement.

pith-pipeline@v1.2.0-daily-grok45 · 27644 in / 4384 out tokens · 73664 ms · 2026-07-30T12:19:29.543309+00:00 · methodology

0 comments
read the original abstract

We develop a structure-preserving neural variational framework for random Young-diagram ensembles, with representations adapted to the structure and scaling of each measure. The method is validated on the Plancherel, uniform, minimal-difference, and fixed-\(q\) \(q\)-Plancherel ensembles, using known asymptotic profiles only for post-training comparison. We then study a quartically deformed hook-length ensemble without assuming an analytical saddle shape. Large-\(n\) neural profiles are compared with finite-size MAP profiles obtained from exact-action searches and with mean profiles obtained from corner-transfer Metropolis--Hastings sampling. Increasing the deformation suppresses the leading rows and broadens the support, while the neural, discrete, and sampled mean profiles agree at the percent level. These results provide numerical evidence for a deformation-dependent macroscopic saddle family.

Figures

Figures reproduced from arXiv: 2607.27061 by Bo-Xuan Ge, Qian Chen.

Figure 1
Figure 1. Figure 1: FIG. 1. Recovery of the Plancherel limit shape from the finite-size soft-hook action at [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Entropy-based recovery of the uniform random-partition profile at [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Minimal-difference partition profiles at [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Fixed- [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Near-unity fixed- [PITH_FULL_IMAGE:figures/full_fig_p022_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. MAP profiles of the quartically deformed hook-length ensemble at [PITH_FULL_IMAGE:figures/full_fig_p022_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Neural profiles for [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Comparison of the neural and MAP profiles for [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. MAP and corner-transfer Metropolis–Hastings mean profiles at [PITH_FULL_IMAGE:figures/full_fig_p024_9.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 3 linked inside Pith

  1. [1]

    The identity X λ⊢n (dimλ) 2 =n! (12) ensures that the measure is normalized

    Plancherel measure The Plancherel measure is defined by PPl n (λ) = (dimλ) 2 n! , λ⊢n,(11) 4 where dimλis the dimension of the irreducible representation ofS n indexed byλ. The identity X λ⊢n (dimλ) 2 =n! (12) ensures that the measure is normalized. By the hook-length formula [1], dimλ= n!Y (i,j)∈λ h(i, j) , h(i, j) =λ i −j+λ ′ j −i+ 1,(13) whereλ ′ is th...

  2. [2]

    No closed-form expression is known, although asymptotic formulas such as the Hardy–Ramanujan formula are available

    Uniform random partitions The uniform ensemble assigns equal probability to every partition ofn: Punif n (λ) = 1 p(n) , λ⊢n,(16) where p(n) = |Yn| denotes the partition number. No closed-form expression is known, although asymptotic formulas such as the Hardy–Ramanujan formula are available. Since all partitions have the same finite-size weight, there is ...

  3. [3]

    , ℓ(λ)−1}.(17) The corresponding ensemble is uniform on this constrained set: P(p) n (λ) = 1 Zn,p 1{λ∈Y(p) n }, Z n,p =|Y (p) n |,(18) where1 A denotes the indicator of the set A

    Minimal-difference-ppartitions Forp∈Z ≥0, define Y(p) n ={λ⊢n:λ i −λ i+1 ≥p, i= 1, . . . , ℓ(λ)−1}.(17) The corresponding ensemble is uniform on this constrained set: P(p) n (λ) = 1 Zn,p 1{λ∈Y(p) n }, Z n,p =|Y (p) n |,(18) where1 A denotes the indicator of the set A. The cases p = 0 and p = 1 give ordinary and distinct partitions, respectively. As p incr...

  4. [4]

    Before evaluating the exact finite-size action, these rows are projected onto an integer partition ofn

    Projection to integer partitions and hard actions The hook-action solvers produce continuous, nonincreasing row lengths. Before evaluating the exact finite-size action, these rows are projected onto an integer partition ofn. We first define eλi = j λ(ϕ) i k , r i =λ (ϕ) i − j λ(ϕ) i k .(A48) Any small numerical violation of monotonicity is removed before ...

  5. [5]

    It serves as the main non-benchmark application of the neural variational solver developed in this work

    Quartically deformed hook-length ensemble We finally introduce a deformed ensemble for which no analytic limit shape is assumed. It serves as the main non-benchmark application of the neural variational solver developed in this work. Forθ≥0 andc h >0, we define the finite-size action S(ch) n,θ (λ) :=c h X u∈λ logh(u) +θ √n X i≥1 λi√n 4 , λ⊢n.(27) The corr...

  6. [6]

    The neural density is optimized using the finite- grid Bose-type entropy introduced in Sec

    Uniform random partitions We first consider ordinary uniform partitions. The neural density is optimized using the finite- grid Bose-type entropy introduced in Sec. III C. Neither the classical limit shape nor the finite-grid entropy maximizer is used during training or checkpoint selection. The calculation is performed at n = 105. The grid begins at xmin...

  7. [7]

    The constraint λi −λ i+1 ≥p introduces an increasing degree of exclusion between neighboring row lengths;p= 1 corresponds to partitions into distinct parts

    Minimal-difference partitions We next consider minimal-difference partitions with p = 1, 2, 3, following the ordinary uniform ensemble as the p = 0 case. The constraint λi −λ i+1 ≥p introduces an increasing degree of exclusion between neighboring row lengths;p= 1 corresponds to partitions into distinct parts. 14 FIG. 2. Entropy-based recovery of the unifo...

    2000

  8. [8]

    Quantum Universe Physical Simulation Platform

    The dashed black curve is the VKLS reference for the undeformed case. Faint traces show the finite-size staircases, while the solid curves are coarse-grained representations used for visualization. FIG. 7. Neural profiles for θ = 0, 0.1, 0.5, and 1 at n = 2 × 107. Solid curves show the continuous neural outputs, while the faint staircases show their integ...

  9. [9]

    Linear weights are initialized with Xavier initialization and biases are set to zero

    Network architecture and numerical settings All neural calculations use fully connected coordinate networks with five hidden layers of width 128 and SiLU activation functions. Linear weights are initialized with Xavier initialization and biases are set to zero. The reported calculations use single-precision floating-point arithmetic and random seed 1234. ...

  10. [10]

    Ordinary Plancherel calculation For the ordinary Plancherel benchmark, the retained row coordinates are xi = i√n , i= 1,

    Hook-action neural solvers a. Ordinary Plancherel calculation For the ordinary Plancherel benchmark, the retained row coordinates are xi = i√n , i= 1, . . . , M, M= Xmax √n .(A1) The reported calculation uses n = 105, Xmax = 2.05, and M = 649. The coordinate supplied to the network is linearly mapped to [−1,1]. The network output is converted to monotone ...

    2000

  11. [11]

    Uniform random partitions For uniform random partitions, we use the grid xk = k√n ,∆x= 1√n , k= 1,

    Entropy-based neural solvers a. Uniform random partitions For uniform random partitions, we use the grid xk = k√n ,∆x= 1√n , k= 1, . . . , K, K= Xmax √n .(A25) The reported calculation uses n = 10 5 and Xmax = 10, giving K = 3163. The grid starts at x1 = 1/√n, since the network input contains logxand the continuum profile is singular atx= 0. The two input...

  12. [12]

    Discrete MAP search and cross-ncontinuation The discrete calculation searches for low-action integer partitions using the exact quartic action in Eq. (A55). Each trial move transfers one box while preserving the total size of the partition. The source row is selected from a mixture of two proposals. With probability 0 .55, row i is chosen with probability...

    2000

  13. [13]

    A rowicontains a removable corner when λi > λi+1,(A60) where the row below the final nonzero row is assigned length zero

    Corner-transfer Metropolis–Hastings sampling The MCMC calculation samples the finite-size quartic ensemble Pn,θ(λ)∝exp h −S(1) n,θ(λ) i (A59) at fixed temperatureT= 1. A rowicontains a removable corner when λi > λi+1,(A60) where the row below the final nonzero row is assigned length zero. The set of removable corners is denoted by R(λ). A proposal first s...

    2000

  14. [14]

    Fulton and J

    W. Fulton and J. Harris,Representation Theory: A First Course, Graduate texts in mathematics (Springer, 1991)

    1991

  15. [15]

    Bacon, I

    D. Bacon, I. L. Chuang, and A. W. Harrow, arXiv preprint quant-ph/0601001 (2005)

    Pith/arXiv arXiv 2005

  16. [16]

    B. F. Logan and L. A. Shepp, Advances in mathematics26, 206 (1977)

    1977

  17. [17]

    S. V. K. A. M. Vershik, Dokl. Akad. Nauk SSSR233, 1024 (1977)

    1977

  18. [18]

    A. M. Vershik and S. V. Kerov, Functional Analysis and Its Applications19, 21 (1985)

    1985

  19. [19]

    Mkrtchyan, European Journal of Combinatorics33, 1631 (2012), groups, Graphs, and Languages

    S. Mkrtchyan, European Journal of Combinatorics33, 1631 (2012), groups, Graphs, and Languages

    2012

  20. [20]

    Borodin, A

    A. Borodin, A. Okounkov, and G. Olshanski, Journal of the American Mathematical Society13, 481 (2000), arXiv:math/9905032 [math.CO]

    Pith/arXiv arXiv 2000

  21. [21]

    W. E and B. Yu, Communications in Mathematics and Statistics6, 1 (2018)

    2018

  22. [22]

    A. M. Vershik, Functional Analysis and Its Applications30, 90 (1996)

    1996

  23. [23]

    Comtet, S

    A. Comtet, S. N. Majumdar, S. Ouvry, and S. Sabhapandit, Journal of Statistical Mechanics: Theory and Experiment , P10001 (2007)

    2007

  24. [24]

    Comtet, S

    A. Comtet, S. N. Majumdar, and S. Sabhapandit, Journal of Mathematical Physics, Analysis, Geometry 4, 24 (2008)

    2008

  25. [25]

    F´ eray and P.-L

    V. F´ eray and P.-L. M´ eliot, Probability Theory and Related Fields152, 589 (2012)

    2012

  26. [26]

    Asymptotics of the gelfand models of the symmetric groups,

    P.-L. M´ eliot, “Asymptotics of the gelfand models of the symmetric groups,” (2010), arXiv:1009.4047 [math.RT]

    Pith/arXiv arXiv 2010

  27. [27]

    Metropolis, A

    N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, The Journal of Chemical Physics21, 1087 (1953)

    1953

  28. [28]

    W. K. Hastings, Biometrika57, 97 (1970)

    1970