Pith. sign in

REVIEW 4 major objections 3 minor 52 references

Nonlinear exchange dynamics that preserve density or marginals converge to the hard-core Gibbs measure, and a particle-system version of them samples independently of unknown fugacities in almost linear time.

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 · deepseek-v4-flash

2026-08-03 15:17 UTC pith:COYGNCTI

load-bearing objection Prescribed-density sampler and subcritical entropy decay are solid and the coupling technique is genuinely new; the prescribed-marginal sampler is conditional on a plausible unproved coloring conjecture that the abstract doesn't flag. the 4 major comments →

arxiv 2607.29016 v1 pith:COYGNCTI submitted 2026-07-31 cs.DS

Nonlinear Exchange Dynamics for Independent Sets

Mehrad Abbaszadeh Minab , Pietro Caputo , Zongchen Chen , Mario Morellini , Alistair Sinclair This is my paper
classification cs.DS MSC 60J1068W2082B20
keywords hard-core modelindependent setsnonlinear dynamicsmean-field dynamicssingle-site dynamicsparticle systemsprescribed density samplingentropy decay
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.

The paper claims that two nonlinear, collision-based dynamics over independent sets—one preserving total density, one preserving per-vertex marginals—converge to the hard-core Gibbs measure with the prescribed conserved quantity, even though fugacity parameters are never estimated. It proves almost linear convergence at sufficiently low density via novel one-step Wasserstein couplings, and exponential relative-entropy decay for the mean-field dynamics up to the critical density. It then shows that a related linear particle-system chain approximates these dynamics and leads to sampling algorithms, most notably an O~(n/epsilon) sampler for fixed-density independent sets below the critical density. If correct, this gives a parameter-free route to maximum-entropy sampling that bypasses the usual inverse-problem step of learning fugacities.

Core claim

The central claim is that the map taking a distribution p to p∘p, where p∘p is the law of the first output of a pairwise exchange collision, converges to the unique hard-core Gibbs measure whose density (mean-field kernel) or marginal vector (single-site kernel) matches that of p. Under low-density assumptions, the paper proves a contraction of the Hamming Wasserstein distance by factor 1 - delta/n per step, yielding O(n log(n/epsilon)) convergence in total variation. For the continuous-time mean-field evolution, it proves exponential decay of relative entropy for all densities up to the critical density alpha_c(Delta), by transferring a modified log-Sobolev inequality from a finite particle

What carries the argument

The central objects are the mean-field and single-site exchange kernels: two independent sets are coupled by swapping bits at randomly chosen vertices, accepted only if both resulting sets remain independent. The mean-field kernel preserves total density; the single-site kernel preserves the vector of per-vertex marginals. Reversibility of the corresponding hard-core Gibbs measures makes these measures stationary, while the particle-system chain—a linear Markov chain on N copies of the graph—approximates the nonlinear evolution and is the executable version of the dynamics. The proof machinery combines a lifting coupling that uses the equality of conserved quantities, an entropy-contraction

Load-bearing premise

The prescribed-marginal sampler relies on an unproved conjecture that Glauber dynamics for proper q-colorings mixes in nearly linear time whenever q >= c_col Delta; if that conjecture fails, the marginal sampler loses its proven mixing guarantee.

What would settle it

Compute the one-step Wasserstein contraction of the mean-field or single-site dynamics on a small graph (for example, a path or cycle) at density near 1/(3(Delta+1)): if a pair of equal-density distributions can be found whose Hamming Wasserstein distance does not shrink by at least the claimed factor, the contraction lemmas fail. Alternatively, a counterexample to the q-coloring Glauber mixing conjecture would directly falsify the fixed-marginal sampler's guarantee.

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

If this is right

  • A new parameter-free algorithm samples fixed-density hard-core independent sets in time O~(n/epsilon) whenever the density is below alpha_c(Delta), improving on the O~(n/epsilon^2) fugacity-learning approach.
  • A fixed-marginal sampler runs in O~(n^3/epsilon^4), with a guarantee that would reach marginals up to 1/(Delta+1) under the widely believed coloring Glauber conjecture.
  • The mean-field dynamics exhibit modified log-Sobolev behavior: relative entropy decays exponentially at rate r/n for all subcritical densities, a nonlinear analog of rapid mixing.
  • Above the critical density alpha_c(Delta), the paper shows that no polynomial-time sampler can exist unless NP=RP, so the subcritical regime is computationally sharp.
  • Both samplers are simple to implement, requiring only direct simulation of a particle system with no parameter learning or inverse-problem subroutines.

Where Pith is reading between the lines

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

  • The entropy-contraction route suggests a general template: for any hard-constrained spin model with a conserved quantity and a reversible Gibbs stationary measure, one may obtain a maximum-entropy sampler by combining a small particle system with a mixing bound for the corresponding particle chain.
  • The low-density threshold 1/(3(Delta+1)) in the quantitative convergence theorems is likely an artifact of the coupling argument; the paper itself conjectures convergence up to 1/(Delta+1), and a sharper coupling would extend both the dynamics and the single-site sampler.
  • A concrete testable extension is to check numerically whether the mean-field particle system's mixing time degrades sharply as the density crosses alpha_c(Delta); the paper's hardness argument predicts such a transition, but a direct finite-N observation would be a useful practical validation.
  • The single-site sampler's coloring-based proof is a placeholder: if the underlying coloring conjecture fails, the algorithm may still mix on the auxiliary graphs that arise here, and a direct mixing proof on those highly structured graphs would restore the guarantee without the conjecture.

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

4 major / 3 minor

Summary. The paper studies nonlinear exchange dynamics for the hard-core model on graphs of maximum degree Δ. Two dynamics are considered: a mean-field kernel preserving density, and a single-site kernel preserving the vertex marginal vector. The paper establishes qualitative convergence to the hard-core Gibbs measure with the prescribed conserved quantity (Propositions 3.5 and 3.7), quantitative Wasserstein contraction at low density (Theorems 4.1 and 4.3), and exponential decay of relative entropy for the continuous-time mean-field dynamics for all subcritical densities (Theorem 4.11). It then presents particle-system samplers: an O~(n/ε) algorithm for prescribed density α≤(1−δ)α_c(Δ) (Theorem 5.4), a hardness result above α_c(Δ) (Theorem 5.6), and an O~(n^3/ε^4) algorithm for prescribed marginals under a coloring-mixing conjecture (Theorem 5.13). Appendices give comparison algorithms based on bisection and gradient descent.

Significance. If the main theorems hold, this is a substantial contribution: it provides one of the first systematic quantitative analyses of nonlinear exchange dynamics for a hard-constraint spin system, introduces coupling arguments that avoid path coupling for linear chains, and demonstrates how Kac's program yields entropy decay in a discrete setting. The density sampler improves the ε-dependence over a bisection-based inverse-problem approach, and the entropy-decay result sharpens the known regime for mean-field dynamics. The paper is also honest in labeling Assumption 4.8 as a conjecture, although the abstract and informal theorem statements do not carry that caveat. The central coupling computations in Lemmas 4.2 and 4.4 appear algebraically correct, and the reduction from the single-site particle chain to Glauber dynamics on an auxiliary coloring graph is elegant and well executed.

major comments (4)
  1. [§4.4.2, Eqs. (61)–(62)] The proof of the entropy-decay theorem is not self-contained at two load-bearing points. The 'entropic chaos' limit (61) is asserted to follow 'by adapting the arguments in [CM25, Proposition 4.9]' and the 'Fisher chaos' limit (62) is asserted 'as in [CM25, Proposition 4.10]', but no adaptation is supplied. These are nontrivial limits for the conditioned product measure on Ω^mf_{N,α_N}; the conditioning event is the fixed-size independent-set space, and the argument must be checked for the specific graph G_N. Similarly, in Theorem 4.6 the LSI (45) is imported from [JMPV23, JPV26] via an unstated 'simple comparison' of the Dirichlet form of P^mf_N with that of the down-up walk; this comparison is not immediate because P^mf_N performs simultaneous two-site swaps. Without these steps, Theorem 4.11 and the density sampler's mixing bound (Lemma 5.3) are not fully proven in this manuscript. Pl
  2. [§5.3 / Assumption 4.8 / Theorem 5.13] The prescribed-marginal sampler is conditional on an unproved conjecture: Assumption 4.8, that Glauber dynamics for q-colorings mixes in near-linear time whenever q ≥ c_col Δ for an absolute constant c_col > 1. The formal Theorem 5.13 therefore only yields the range m_v ≤ 1/(c_col(Δ+1)), with c_col unknown. This is flagged in the body, but the abstract and the informal Theorem 1.5 in the introduction claim 'sampling with a specified marginal vector' with no such qualification. Since this assumption is the only bridge from the coloring chain to the single-site particle chain in Theorem 4.10, the second headline algorithmic contribution is conditional. The paper should state this conditionality in the abstract and Section 1, and should explicitly mention what is known unconditionally (e.g., Jerrum's c_col > 2 bound gives m_v ≤ 1/((2+ε)(Δ+1))).
  3. [§1.3 vs §5.3] The informal statement of the marginal sampler in Theorem 1.5 gives the upper bound m_v ≤ (1−δ)/(2(Δ+1)) and does not mention Assumption 4.8, while the formal Theorem 5.13 gives m_v ≤ 1/(c_col(Δ+1)) under Assumption 4.8. These are not consistent: the informal bound corresponds to a specific numerical value of c_col (around 2), not to the unknown c_col from the conjecture. The introduction should either state the theorem as conditional exactly as in Section 5.3 or openly say it is an informal consequence for the conjectured c_col ≈ 1. As written, the informal statement overstates the proven result.
  4. [§5.3, Lemma 5.11 and Corollary 5.12] There is a technical inconsistency in the definition of L. Lemma 5.11 defines L = √(N/2) · log(4n), but the subsequent Hoeffding step '2 exp(−2L^2/N) = 2e^{−log(4n)}' is correct only for L = √(N/2 log(4n)). With the displayed L, the sufficient condition in Corollary 5.12 should be N ≥ C n^2 log^2(4n)/ε^4 rather than N ≥ C n^2 log(4n)/ε^4. The proof of Theorem 5.13 relies on the smaller N. This is fixable by redefining L (or adjusting the polynomial in Corollary 5.12) and does not change the O~(n^3/ε^4) running time, but the proof as written is not correct.
minor comments (3)
  1. [§3.3, Lemma 3.4, Eq. (28)] The final lower bound in Step 3 is missing a factor n/δ: the displayed n! (αδ/n^2)^n (Δ+δ)/(Δ+1) is weaker than the derived n! (αδ/n^2)^n (n/δ)(Δ+δ)/(Δ+1). The displayed bound is still true and the argument is unaffected, but the equation should be corrected.
  2. [§4.4, Definition 4.12] The notation 'Ent_µ[f] = µ[f log(f/logµ[f])]' appears to contain a typo; it should be 'f log(f/µ[f])' (or the standard form Ent_µ[f] = µ[f log f] − µ[f] log µ[f]).
  3. [§5.1, Algorithm 4 / Theorem 5.4] The running time is stated as O_{γ,δ,Δ}(n/ε log(n/ε)) in the proof but O~(n/ε) in the theorem; this is fine as the tilde is defined, but for clarity the proof should refer to the tilde notation.

Circularity Check

0 steps flagged

No circular derivation; central results are built from internal couplings and independent prior theorems, with minor self-citations and one explicitly conjectural coloring-mixing assumption.

full rationale

I walked the derivation chain and found no step in which a claimed prediction is equivalent by construction to an input, or in which a fit is renamed as a prediction. The qualitative convergence results (Propositions 3.5 and 3.7) invoke [CS24, Theorem 2.8] and [CS24, Proposition 2.10] as black boxes; these are prior published theorems with general hypotheses (balance, positive diagonal, irreducibility) that do not assume the hard-core conclusions. The hard-core-specific content is proven in the paper: conservation laws (Lemma 2.1), reversibility of the Gibbs measures (Lemma 2.2), irreducibility (Lemmas 3.4 and 3.6), and identification of stationary measures (Lemma 3.3). The quantitative Wasserstein contraction (Theorems 4.1 and 4.3) is obtained through explicit couplings (Algorithms 1 and 2); the contraction factor follows from the assumed density/marginal bounds and is not fitted to the conclusion. The entropy decay proof (Theorem 4.11) uses the particle-system LSI from [JMPV23, JPV26] and chaos-transfer propositions from [CM25]; these are independent prior results with stated assumptions that do not include the target inequality, so the self-citation is not load-bearing in a circular sense. The single-site sampler (Theorem 5.13) rests on Assumption 4.8, which the paper explicitly labels as a widely believed conjecture; this is a conditionality/correctness risk, not an input-output identification. No self-definitional, fitted-input, ansatz-smuggling, or renaming step was found. The self-citations are real evidence because the cited statements are general mathematical theorems, not restatements of the current claims.

Axiom & Free-Parameter Ledger

1 free parameters · 5 axioms · 0 invented entities

No empirical fitting is present; constants are existential. The paper depends on several external theorems and on one explicitly flagged conjecture. The authors' own prior framework is used as black-box support rather than as a restatement of the current claims.

free parameters (1)
  • c_col
    Unknown absolute constant in Assumption 4.8, assumed to exist for the coloring mixing conjecture. Theorem 5.13 and the marginal sampler depend on it, but no value is supplied.
axioms (5)
  • domain assumption Entropy contraction / log-Sobolev inequality for the mean-field particle system at subcritical densities (Theorem 4.6)
    Imported from [JMPV23, Theorem 2] and [JPV26, Equation 6.3]. It underpins the entropy decay of the nonlinear dynamics and the mixing of the density sampler.
  • ad hoc to paper Coloring Glauber dynamics mixes in near-linear time for q ≥ c_col Δ (Assumption 4.8)
    Unproved, widely believed conjecture. Used in Theorem 4.10 and Theorem 5.13 to bound the mixing of the single-site particle system.
  • standard math General convergence criterion for nonlinear dynamics [CS24, Theorem 2.8]
    Black-box theorem from the authors' prior work; used for qualitative convergence in Propositions 3.5 and 3.7.
  • standard math Local central limit theorem and characteristic-function estimates [JMPV23, JPSS22]
    Used in Lemma 5.2 to control the one-particle marginal of the conditioned mean-field particle system.
  • standard math Complete spectral independence and Glauber rapid mixing below the uniqueness threshold [CFYZ21, CLV23, AJK+22, CFYZ22, CE25]
    Used in Appendix B for smoothness of the covariance and for sampling at subcritical fugacities in the gradient/bisection algorithms.

pith-pipeline@v1.3.0-daily-deepseek · 67669 in / 21217 out tokens · 211738 ms · 2026-08-03T15:17:04.345863+00:00 · methodology

0 comments
read the original abstract

In recent years, nonlinear dynamics derived from kinetic theory have gained attention in the context of sampling configurations of spin systems such as the Ising model. We focus on nonlinear dynamics for the hard-core model, a canonical spin system with hard constraints that specifies a distribution over independent sets in a graph, weighted by their sizes. We explore two distinct types of nonlinear dynamics: the mean-field dynamics, which preserves the density (or average size) of independent sets, and the single-site dynamics, which preserves the marginal vector (i.e., the occupancy probabilities of the vertices). These dynamics are natural stochastic processes for sampling from the hard-core model with a specified density or marginal vector, respectively, both of which are canonical instances of maximum entropy distributions that have been studied in various contexts. In contrast to linear Markov chains, there is a significant lack of a fundamental theoretical framework for nonlinear dynamics. We develop foundational theoretical tools for analyzing nonlinear dynamics within the context of the hard-core model. We establish almost linear convergence of both the mean-field and single-site dynamics at sufficiently low density through novel coupling arguments. We also establish exponential decay of relative entropy for the mean-field dynamics all the way up to the critical density. Additionally, we design new algorithms for sampling from the hard-core distribution with either a specified density or a specified marginal vector. These algorithms are based on a related linear Markov chain, called the particle-system dynamics and inspired by the so-called Kac's program, that approximates the associated nonlinear dynamics. As we demonstrate in the paper, they are comparable in time complexity, but simpler to implement, than traditional approaches based on learning parameter values.

Figures

Figures reproduced from arXiv: 2607.29016 by Alistair Sinclair, Mario Morellini, Mehrad Abbaszadeh Minab, Pietro Caputo, Zongchen Chen.

Figure 1
Figure 1. Figure 1 [PITH_FULL_IMAGE:figures/full_fig_p005_1.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

52 extracted references · 1 canonical work pages

  1. [1]

    2019 , publisher =

    Foundations of Chemical Reaction Network Theory , author =. 2019 , publisher =

    2019

  2. [2]

    , title=

    Villani, C. , title=. Handbook of Mathematical Fluid Dynamics , volume=

  3. [3]

    , title=

    Villani, C. , title=. Entropy Methods for the Boltzmann Equation , series=

  4. [4]

    , title=

    Weinberg, W. , title=. Naturkunde in W\"urttemberg , volume=

  5. [5]

    , title=

    Hardy, G.H. , title=. Science , volume=

  6. [6]

    and Jasra, A

    Andrieu, C. and Jasra, A. and Doucet, A. and Del Moral, P. , title=. Bernoulli , volume=

  7. [7]

    , title=

    Butkovsky, O. , title=. Theory of Probability & Its Applications , volume=

  8. [8]

    , title=

    Mitchell, M. , title=

  9. [9]

    Path coupling:

    Bubley, Russ and Dyer, Martin , booktitle=. Path coupling:. 1997 , organization=

    1997

  10. [10]

    The Annals of Applied Probability , volume=

    Cutoff phenomenon in nonlinear recombinations , author=. The Annals of Applied Probability , volume=. 2025 , publisher=

    2025

  11. [11]

    Kinetic and Related Models , volume=

    Entropy and chaos in the Kac model , author=. Kinetic and Related Models , volume=. 2010 , publisher=

    2010

  12. [12]

    Inventiones mathematicae , volume=

    Kac’s program in kinetic theory , author=. Inventiones mathematicae , volume=. 2013 , publisher=

    2013

  13. [13]

    Cercignani, Carlo , year=. The

  14. [14]

    , title=

    Goldberg, D.E. , title=

  15. [15]

    Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, Volume 3: Contributions to Astronomy and Physics , pages=

    Foundations of Kinetic Theory , author=. Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, Volume 3: Contributions to Astronomy and Physics , pages=. 1956 , publisher=

    1956

  16. [16]

    Entropy production in nonlinear recombination models , volume =

    Pietro Caputo and Alistair Sinclair , journal =. Entropy production in nonlinear recombination models , volume =

  17. [17]

    Proceedings of the 33rd Annual IEEE Symposium on Foundations of Computer Science (FOCS 1992) , pages=

    Quadratic dynamical systems , author=. Proceedings of the 33rd Annual IEEE Symposium on Foundations of Computer Science (FOCS 1992) , pages=. 1992 , organization=

    1992

  18. [18]

    and Rabinovich, Y

    Rabani, Y. and Rabinovich, Y. and Sinclair, A. , title =. Random Structures & Algorithms , volume =. doi:https://doi.org/10.1002/(SICI)1098-2418(199807)12:4<313::AID-RSA1>3.0.CO;2-W , year =

    work page doi:10.1002/(sici)1098-2418(199807)12:4

  19. [19]

    and Saberi, A

    Oveis-Gharan, S. and Saberi, A. and Singh, M. , title=. Proceedings of the 52nd Annual IEEE Symposium on Foundations of Computer Science , pages =. 2011 , publisher =

    2011

  20. [20]

    and Goemans, M

    Asadpour, A. and Goemans, M. and Madry, A. and Oveis-Gharan, S. and Saberi, A. , title=. Proceedings of the 21st Annual ACM-SIAM Symposium on Discrete Algorithms , pages =. 2010 , publisher =

    2010

  21. [21]

    , title=

    Kapur, J.N. , title=

  22. [22]

    , title=

    Jaynes, E.T. , title=. Physical Review , volume=

  23. [23]

    , title =

    Singh, Mohit and Vishnoi, Nisheeth K. , title =. 2014 , publisher =. doi:10.1145/2591796.2591803 , booktitle =

    arXiv 2014

  24. [24]

    Nguyen, H. C. and Zecchina, R. and Berg, J , year =. Inverse statistical problems: from the inverse. Advances in Physics , volume=

  25. [25]

    Nonlinear dynamics for the

    Caputo, Pietro and Sinclair, Alistair , journal =. Nonlinear dynamics for the. 2024 , volume =

    2024

  26. [26]

    Laurent Miclo , title =. S

  27. [27]

    2023 , doi =

    Ewan Davies and Will Perkins , title =. 2023 , doi =

    2023

  28. [28]

    2022 , doi =

    Vishesh Jain and Will Perkins and Ashwin Sah and Mehtaab Sawhney , title =. 2022 , doi =

    2022

  29. [29]

    and Jordan, Michael I

    Wainwright, Martin J. and Jordan, Michael I. , title =. Foundations and Trends in Machine Learning , volume =

  30. [30]

    and Stewart, Alistair and Sun, Yuxin , title =

    Diakonikolas, Ilias and Kane, Daniel M. and Stewart, Alistair and Sun, Yuxin , title =. arXiv preprint arXiv:2102.02171 , year =

    Pith/arXiv arXiv

  31. [31]

    arXiv preprint arXiv:2105.15005 , year =

    Chen, Xiaoyu and Feng, Weiming and Yin, Yitong and Zhang, Xinyuan , title =. arXiv preprint arXiv:2105.15005 , year =

    Pith/arXiv arXiv

  32. [32]

    2024 , eprint=

    Flip Dynamics for Sampling Colorings: Improving (11/6- ) Using a Simple Metric , author=. 2024 , eprint=

    2024

  33. [33]

    arXiv preprint arXiv:2511.05223 , year=

    Kac's Program and Relative Entropy Decay for Nonlinear Spin-Exchange Dynamics , author=. arXiv preprint arXiv:2511.05223 , year=

    arXiv

  34. [34]

    2023 , note =

    Caputo, Pietro , title =. 2023 , note =

    2023

  35. [35]

    arXiv preprint arXiv:2604.10902 , year=

    Entropic independence via sparse localization , author=. arXiv preprint arXiv:2604.10902 , year=

    Pith/arXiv arXiv

  36. [36]

    Optimal Mixing of the Down-Up Walk on Independent Sets of a Given Size , booktitle =

    Vishesh Jain and Marcus Michelen and Huy Tuan Pham and Thuy. Optimal Mixing of the Down-Up Walk on Independent Sets of a Given Size , booktitle =. 2023 , doi =

    2023

  37. [37]

    2022 IEEE 63rd Annual Symposium on Foundations of Computer Science (FOCS) , pages=

    Optimal mixing for two-state anti-ferromagnetic spin systems , author=. 2022 IEEE 63rd Annual Symposium on Foundations of Computer Science (FOCS) , pages=. 2022 , organization=

    2022

  38. [38]

    Localization schemes:

    Chen, Yuansi and Eldan, Ronen , journal=. Localization schemes:. 2025 , publisher=

    2025

  39. [39]

    Proceedings of the 57th Annual ACM Symposium on Theory of Computing , pages=

    Rapid mixing at the uniqueness threshold , author=. Proceedings of the 57th Annual ACM Symposium on Theory of Computing , pages=

  40. [40]

    Probability Theory and Related Fields , volume=

    On the hardness of sampling independent sets beyond the tree threshold , author=. Probability Theory and Related Fields , volume=. 2009 , publisher=

    2009

  41. [41]

    Random Structures & Algorithms , volume =

    Jerrum, Mark , title =. Random Structures & Algorithms , volume =

  42. [42]

    Proceedings of the thirty-eighth annual ACM symposium on Theory of computing , pages=

    Counting independent sets up to the tree threshold , author=. Proceedings of the thirty-eighth annual ACM symposium on Theory of computing , pages=

  43. [43]

    SIAM Journal on Computing , volume=

    Spectral independence in high-dimensional expanders and applications to the hardcore model , author=. SIAM Journal on Computing , volume=. 2021 , publisher=

    2021

  44. [44]

    Asadpour, Arash and Goemans, Michel X and Madry, Aleksander and. An. Operations Research , volume=. 2017 , publisher=

    2017

  45. [45]

    Maximum entropy distributions:

    Straszak, Damian and Vishnoi, Nisheeth K , booktitle=. Maximum entropy distributions:. 2019 , organization=

    2019

  46. [46]

    Data preprocessing to mitigate bias:

    Celis, L Elisa and Keswani, Vijay and Vishnoi, Nisheeth , booktitle=. Data preprocessing to mitigate bias:. 2020 , organization=

    2020

  47. [47]

    IEEE Transactions on Information Theory , volume=

    Generalized maximum entropy for supervised classification , author=. IEEE Transactions on Information Theory , volume=. 2022 , publisher=

    2022

  48. [48]

    Electronic Journal of Statistics , volume=

    Computational implications of reducing data to sufficient statistics , author=. Electronic Journal of Statistics , volume=

  49. [49]

    Advances in Neural Information Processing Systems , volume=

    Hardness of parameter estimation in graphical models , author=. Advances in Neural Information Processing Systems , volume=

  50. [50]

    Optimal Mixing of

    Chen, Zongchen and Liu, Kuikui and Vigoda, Eric , journal=. Optimal Mixing of. 2023 , publisher=

    2023

  51. [51]

    Entropic independence:

    Anari, Nima and Jain, Vishesh and Koehler, Frederic and Pham, Huy Tuan and Vuong, Thuy-Duong , booktitle=. Entropic independence:

  52. [52]

    The Computational Hardness of Counting in Two-Spin Models on d-Regular Graphs , year=

    Sly, Allan and Sun, Nike , booktitle=. The Computational Hardness of Counting in Two-Spin Models on d-Regular Graphs , year=