Pith. sign in

REVIEW 3 major objections 4 minor 63 references

Identifying nonequilibrium degrees of freedom in high-dimensional stochastic systems

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A neural network whose logit is a quadratic form of learned embeddings isolates the few coordinates that actually produce entropy, and estimates the entropy production rate from raw time series at the same time.

desk verdict A solid method paper for linear systems whose nonlinear identification claim rests on an unproven identifiability assumption. read the letter →

arxiv 2508.08247 v2 pith:SVCSV5YK submitted 2025-08-11 cond-mat.stat-mech physics.bio-ph

Catherine Ji , Ravin Raj , Benjamin Eysenbach , Gautam Reddy This is my paper · ORCID
classification cond-mat.stat-mechphysics.bio-ph PACS 05.70.Ln05.40.-a05.10.Gg
keywords entropyproductionnonequilibriumsteadystateirreversibledegreesoffreedomrepresentationlearningcontrastivelow-rankapproximationstochasticdynamicsdimensionalityreduction
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

LENS (Latent Embeddings of Nonequilibrium Systems) is a model-free method that learns a low-dimensional representation of a high-dimensional stochastic system in a nonequilibrium steady state, such that the representation exposes the degrees of freedom that generate entropy production. The method trains a classifier to distinguish forward from time-reversed transitions, but instead of using a free-form logit, it constrains the logit to a specific quadratic form built on learned embeddings. This constraint makes the learned representation interpretable as the irreversible flow coordinates, and the classifier's average logit directly estimates the entropy production rate. The paper demonstrates on simulated linear and nonlinear systems that LENS both identifies circulating flows in the learned representation space and estimates entropy production rates with small error, outperforming PCA-based projections.

What carries the argument

The central object is the LENS logit $S_\theta(\vec x,\vec x')$, a quadratic form on the learned representation $\vec\phi_\theta(\vec x)$, antisymmetric under exchange of its arguments by construction. The structure combines an antisymmetric bilinear term $\vec\phi^T A \vec\phi$ (which captures the circulating, entropy-producing flow) with a difference of symmetric quadratic forms (which captures the potential-like, reversible part). This parametrization is universal for approximating any antisymmetric logit (Theorem 1 and Appendix B3) and yields a precise low-rank geometric interpretation for linear systems: maximizing the LENS objective leads to learning $\tilde K + \tilde L = P^T(K+L)P \approx \Lambda = D^{-1}G + C^{-1}$, the thermodynamic force matrix. The EPR is then the expectation of the logit, $\dot S = \mathbb E[S_\theta(\vec x,\vec x')]$, which is directly computable from the classifier output.

What would settle it

A concrete falsifier would be a simulation of a nonlinear stochastic system with known latent linear irreversible coordinates (e.g., two latent modes undergoing a circulating flow, observed through a highly nonlinear invertible map). If LENS, trained with $M=2$, yields embeddings that are not related to the true latent coordinates by any smooth invertible transformation, or if the estimated EPR from the learned representation fails to reproduce the analytic EPR despite the classifier achieving low validation loss, then the identification claim would be refuted. Another check: for a system where the true irreversible coordinates are known a priori, the eigenvalues of the learned $A_\theta$ should match the true circulation strengths up to permutation; if they do not, the learned representation is not uncovering the irreversible degrees of freedom.

Watch

Extended reading notes

Core claim

The central claim is that for a Markovian nonequilibrium steady state, the optimal classifier logit between forward and reverse transition probabilities, $S^*(\vec x,\vec x') = \log[p(\vec x\to\vec x')/p(\vec x'\to\vec x)]$, can be parametrized without loss of expressiveness as $S_\theta(\vec x,\vec x') = \vec\phi_\theta(\vec x')^T A_\theta \vec\phi_\theta(\vec x) + \tfrac12 \vec\phi_\theta(\vec x')^T B_\theta \vec\phi_\theta(\vec x') - \tfrac12 \vec\phi_\theta(\vec x)^T B_\theta \vec\phi_\theta(\vec x)$, where $\vec\phi_\theta$ is a learned nonlinear representation, $A_\theta$ is block-diagonal skew-symmetric, and $B_\theta$ is symmetric. With this parametrization, the learned representations correspond to the irreversible degrees of freedom: for linear systems, LENS learns a low-rank approximation to the thermodynamic force matrix $\Lambda = D^{-1}G + C^{-1}$, so the entropy production rate from the low-dimensional projection converges monotonically to the true EPR as the representation dimension $M$ increases. For nonlinear systems, the quadratic form is universal (proved via Stone-Weierstrass) in approximating any smooth antisymmetric logit, so the same construction extracts entropy-producing coordinates without requiring knowledge of the dynamics.

Load-bearing premise

The paper's identification claim relies on the assumption that, for nonlinear stochastic systems, optimizing the quadratic logit with a learned representation will not merely fit the scalar entropy production but will actually align the learned embeddings with the latent irreversible degrees of freedom; the authors prove universality for approximating any antisymmetric logit and give a precise low-rank interpretation only for linear dynamics, while explicitly stating that a characterization for nonlinear latent linear dynamics is future work.

Editorial extensions

If this is right

  • For linear stochastic systems with unknown parameters, LENS provides a scalable estimator of the entropy production rate that requires no knowledge of the drift or diffusion matrices, and the learned projection $P$ reveals which combinations of coordinates carry the irreversible driving.
  • For nonlinear systems with latent linear irreversible dynamics, the universality result suggests that LENS can discover the latent coordinates that generate entropy, even when the observed high-dimensional dynamics are nonlinear functions of those coordinates.
  • The method is applicable to partially observed data (e.g., rendered images of bead positions, video frames of complex fields) by using convolutional encoders, making it a candidate for estimating dissipation in experimental active matter and biological systems from raw imaging time series.
  • The low-rank interpretation gives a principled way to choose the representation dimension $M$: the relative EPR error decreases monotonically with $M$ and vanishes when $M = N$, providing a diagnostic for how many irreversible degrees of freedom are needed.
  • By identifying which state-space directions carry the entropy production, LENS can serve as a dimensionality-reduction tool that preserves thermodynamic relevance rather than mere reconstruction, complementing or replacing PCA in nonequilibrium studies.

Reading between the lines

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

  • Editorial inference: If LENS embeddings indeed align with the true latent irreversible coordinates in nonlinear systems, the method could be used not only to estimate EPR but also to construct reduced models of the nonequilibrium dynamics, by projecting the learned representation onto the dissipative subspace and integrating that low-dimensional flow.
  • Editorial inference: The universality proof applies to any smooth antisymmetric logit, suggesting that the same quadratic-form parametrization could serve as a universal approximator for other antisymmetric functions arising in stochastic thermodynamics, such as time-asymmetric correlation functions or the full counting statistics generator.
  • Editorial inference: A testable extension is to apply LENS to systems with explicit latent linear irreversible dynamics and a nonlinear observation map (e.g., a chaotic measurement function), to empirically check whether the learned embedding recovers the latent coordinates up to an invertible transformation; the paper states this characterization is future work.
  • Editorial inference: The method's dependence on a binary cross-entropy objective (rather than the f-divergence used by NEEP) appears to provide training stability at high EPR, so a hybrid objective that interpolates between BCE and f-divergence could potentially improve accuracy at intermediate data sizes.
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

3 major / 4 minor

Summary. The paper introduces LENS, a contrastive learning method that parameterizes the forward/reverse classifier logit as a quadratic form of a learned state representation. The average logit estimates the entropy production rate (EPR), while the learned representation is intended to expose the system's irreversible degrees of freedom. The authors prove a universality result for the quadratic logit, derive an exact low-rank interpretation for linear dynamics, and test the method on three systems: coupled harmonic beads, a driven particle in a periodic potential, and the two-dimensional complex Ginzburg-Landau model. Code is provided.

Significance. If the identification claim holds, LENS would be a valuable advance over existing EPR-only estimators because it adds interpretable coordinates to the entropy production estimate. The paper has real strengths: the analytic EPR benchmarks for the N-bead and periodic-potential systems are convincing, the low-rank theory for linear systems in App. F gives a parameter-free prediction that matches the observed scaling with representation dimension, error bars over random seeds are reported, and the code is publicly available. However, the central identification claim is currently established only for linear dynamics. For nonlinear systems the evidence consists of visual flow inspection, a known periodic coordinate inserted by hand as a sine layer, and a deterministic CGL example with no ground-truth latent coordinates; the Discussion and App. A4 explicitly defer the needed characterization. As written, the paper is a solid EPR estimator with a promising but unproven coordinate-discovery component.

major comments (3)
  1. [Discussion and Outlook; App. A4] The paper's central claim is that LENS identifies latent irreversible degrees of freedom in nonlinear systems, but the nonlinear experiments either use a known periodic coordinate inserted by hand (Fig. 3 and App. A5) or are interpreted through visual flow inspection without ground truth (Fig. 4). The Discussion states that 'a mathematical and empirical characterization of LENS in such scenarios is an interesting direction for future work,' and App. A4 says that consistency of phi in nonlinear systems is open. These statements concede exactly the load-bearing part of the identification claim. I request either (i) a benchmark with known latent irreversible coordinates, for example nonlinear observations generated by latent linear dynamics as named in the Discussion, in which the learned phi is compared quantitatively with the ground-truth coordinates (e.g., by canonical-correlation or shared-variance measures), or (ii) a substantial tempering of the identification claim in the title and abstract to reflect that coordinate identification is currently demonstrated for linear systems only.
  2. [Eq. (3); App. B3; App. A4] The logit parameterization is gauge-invariant under phi -> T phi, A -> T^{-T} A T^{-1}, B -> T^{-T} B T^{-1} for any invertible T, so the LENS objective does not select a unique embedding. The Schur block-diagonal form of A (Cor. 1.1) canonicalizes A alone; because B transforms simultaneously and is otherwise unconstrained, the combined gauge is not fixed. App. A4 in fact acknowledges that 'it is not possible to simultaneously block-diagonalize the skew-symmetric matrix A and diagonalize the symmetric matrix B.' Thus the learned coordinates are not identifiable as physical degrees of freedom without additional assumptions. At minimum, the paper should state the equivalence class of representations that the objective identifies and prove or test which member of that class is selected by the optimizer; the eigenvalue reproducibility claim in Fig. S6 concerns only the directly-learned linear projection and does not cover nonlinear phi.
  3. [Fig. 3; App. A5] The driven-periodic-potential experiment does not provide evidence for discovery of an unknown latent coordinate. Because the periodicity is known a priori, a sine activation layer is inserted after the first dense layer and is described as 'necessary for stable maximization' in App. A5. The learned representation is therefore strongly biased toward the known periodic variable rather than discovering it from data. This leaves the EPR accuracy results intact, but it removes this experiment as support for the identification claim.
minor comments (4)
  1. [App. B3, proof of Lemma 2] After Eq. (S44), the text reads 'This implies that A^T = A, so A is skew-symmetric as required'; it should read A^T = -A.
  2. [Fig. 2(a) and main text after Eq. (4)] The main text refers to 'four different temperature profiles' but does not name them at first mention; the profiles are only listed in the Fig. 2 caption. Please cite Fig. 2(a) at that point so the reader can identify the profiles.
  3. [Table S6 and App. A3b] Table S6 lists temperature ratios {0.05, 0.1, 0.25, ..., 0.9999} that appear copied from the N=2 bead system, while the surrounding text describes a sample-efficiency comparison for the N=16 alternate-profile system with a single temperature ratio of 0.1. Please correct the table or clarify which ratios were used for the N=16 experiment.
  4. [Fig. 4(a)] Because the CGL dynamics are deterministic, the text correctly notes that the EPR estimates diverge logarithmically; the y-axis label 'EPR Estimate' may nonetheless mislead readers. Consider labeling this quantity as a finite-time irreversibility proxy and specifying the observation time window used.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: LENS estimates EPR through the Bayes-optimal classifier identity and validates against independent analytic baselines; the representation-identification claim is limited but not circular.

full rationale

The paper's derivation chain is self-contained. The EPR estimator rests on the textbook identity that the Bayes-optimal logit for forward/reverse discrimination equals log[p(x->x')/p(x'->x)] (App. B2, Eq. S35), so averaging the learned logit estimates the KL divergence in Eq. (1). This is a derivation of an estimator, not a circular reuse of the target: the network parameters are fit to data and the resulting EPR is compared with independent analytic solutions for the N-bead (App. C) and tilted periodic potential (App. D) systems. The low-rank result for linear dynamics (App. F, Eq. S4) is obtained by differentiating the reduced objective (5); it is a mathematical consequence of the model, not an input. The block-diagonal gauge (Cor. 1.1) is a normalization choice and is acknowledged not to fix the representation uniquely (App. A4: 'there are an infinite number of choices for P, A, B...'). The only self-citation of note is the JaxGCRL codebase (Ref. [43]) by a coauthor, used for training infrastructure; it does not supply any physics result. The claim that learned embeddings identify latent irreversible degrees of freedom in nonlinear systems is explicitly deferred ('A mathematical and empirical characterization of LENS in such scenarios is an interesting direction for future work'), which is a scope/identifiability limitation rather than circularity. No fitted parameter is renamed as a prediction, and no load-bearing argument reduces to a self-citation.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The method itself is an estimator: A_theta, B_theta, encoder weights, and M are fitted or hand-chosen, and the universality theorem supplies expressiveness rather than parameter-free prediction. Physical conclusions rest on stationarity, the short-time expansion, optimization convergence, and, for CGL, a heuristic interpretation of a divergent EPR. No new physical entities are postulated.

free parameters (4)
  • Representation dimension M
    User-selected hyperparameter controlling the rank of the entropy-production approximation; results in Figs. 2 through 4 depend on M. Not fitted by theory.
  • Learned network weights and matrices A_theta, B_theta = trained values, not reported
    The LENS estimator fits these parameters to trajectory data by maximizing Eq. (2). They are the mechanism of the method rather than a hidden assumption, but the EPR estimates depend on their convergence.
  • Sine activation layer for periodic potential
    Introduced ad hoc because the periodicity of the driven particle system is known a priori; required for stable convergence in the linear network, as shown in Fig. S7.
  • EPR positivity floor of 1e-8 = 1e-8
    Applied to ensemble EPR estimates in Fig. 2b after observing occasional negative estimates near true EPR of order 1e-6. Disclosed in Appendix A2, but the effect on reported accuracy is not quantified.
assumptions (5)
  • domain assumption The observed trajectory is a stationary Markov process in a nonequilibrium steady state.
    Used to define EPR as the KL divergence between forward and reverse transition probabilities (Eq. 1) and to replace ensemble expectations with trajectory averages. Stated in the Introduction and Appendix B.
  • domain assumption Timestep dt is small enough that the target logit expansion to second order in the displacement is accurate.
    Appendix B expands the logit using a short-time Gaussian transition kernel and truncates at O(||delta||^2). Simulations use dt = 1e-2 or 1e-3 with autocorrelation time of order 1.
  • standard math The state space can be treated as compact for the Stone-Weierstrass universality proof.
    Theorem 1 assumes S is defined on a compact set K x K. Physical state spaces such as Gaussian steady states are unbounded, so the universality statement is an approximation.
  • ad hoc to paper The neural network optimization reaches the Bayes-optimal or near-optimal classifier on the given data.
    The EPR estimator is the average learned logit; any optimization gap or finite-sample error biases the estimate. Convergence is monitored by early stopping, but no global-optimality guarantee is provided.
  • ad hoc to paper For deterministic CGL dynamics, the learned logit can be interpreted as a finite-time irreversibility proxy even though the true EPR diverges logarithmically.
    The paper uses LENS estimates as transition indicators in Fig. 4 while noting that the true EPR diverges for deterministic forward-only dynamics. The proxy is not derived.
invented entities (1)
  • Learned representation phi_theta(x)
    purpose: Defines the low-dimensional subspace in which LENS places irreversible flows and local entropy production.
    The representation is internal to the network and carries gauge freedom. Its physical alignment with latent irreversible coordinates is not independently measured for nonlinear systems.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Identifying nonequilibrium degrees of freedom in high-dimensional stochastic systems." pith.science (2026). https://pith.science/paper/SVCSV5YK

@misc{pith2026250808247,
  author       = {Pith},
  title        = {Pith review of: Identifying nonequilibrium degrees of freedom in high-dimensional stochastic systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SVCSV5YK}},
  note         = {Machine review of arXiv:2508.08247}
}
read the original abstract

Any coarse-grained description of a nonequilibrium system should faithfully represent its latent irreversible degrees of freedom. However, standard dimensionality reduction methods typically prioritize accurate reconstruction over physical relevance. Here, we introduce a model-free approach to identify irreversible degrees of freedom in stochastic systems that are in a nonequilibrium steady state. Our method leverages the insight that a black-box classifier, trained to differentiate between forward and time-reversed trajectories, implicitly estimates the local entropy production rate. By parameterizing this classifier as a quadratic form of learned state representations, we obtain nonlinear embeddings of high-dimensional state-space dynamics, which we term Latent Embeddings of Nonequilibrium Systems (LENS). LENS effectively identifies low-dimensional irreversible flows and provides a scalable, learning-based strategy for estimating entropy production rates directly from high-dimensional time series data.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 48 canonical work pages

  1. [1]

    Decomposition of Stochastic Dynamics Consider a stationary Markovian system with transi- tion probability density p(⃗x→⃗x′), which represents the probability that the system exhibits a transition from state⃗x to a new state ⃗x′ at consecutive time steps t and t + 1, respectively. This joint probability can be equiv- alently written as p(st = ⃗x,st+1 = ⃗x′...

    work page

  2. [2]

    early stopping patience

    N-Beads Model The dynamics of the N-bead system can be described by the linear overdamped Langevin equation ⃗xt+dt =⃗xt + G⃗xt dt + F⃗ξt √ dt (S1) where the deterministic coupling matrix G has elements defined byGij = (−2δi,j+δi,j+1+δi,j−1)k/γ, and the ma- trix F sets the local thermal noise scale. Here, we assume that all springs are identical and that e...

    work page

  3. [3]

    Comparisons between Binary Cross Entropy (BCE) and Variational f-Divergence Objectives A core contribution of LENS is the parameterization of EP as a function of state representations, where the local (one-step) EP takes a quadratic form characteristic of linear stochastic dynamical systems (see Sec. B 1 a). However, the question remains on the efficacy o...

    work page

  4. [4]

    Directly learning (P,⃗b) eliminates the need for a large network and sig- nificantly reduces parameter count

    Directly Learning Projection Matrices for Linear Stochastic Dynamics For linear systems, rather than learning a transforma- tion of inputs via a linear neural network, we can directly fit an affine mapping ⃗ϕ(⃗x) = P⃗x +⃗b. Directly learning (P,⃗b) eliminates the need for a large network and sig- nificantly reduces parameter count. Furthermore, this param...

    work page

  5. [5]

    Simulation and Training Hyperparameters of the N = 16 Bead System for Directly-Learned Lin- ear Projections Hyperparameter values correspond to those used for data plotted in Fig

    Driven Overdamped Particle in a Periodic Potential The dynamics of an overdamped particle in a periodic potential being driven by a constant force h is given by xt+1 =xt + h− dU dx dt + √ 2D dtξt, (S2) TABLE S7. Simulation and Training Hyperparameters of the N = 16 Bead System for Directly-Learned Lin- ear Projections Hyperparameter values correspond to t...

    work page

  6. [6]

    Here, α and β are the linear and nonlinear dispersions of the medium respectively

    2D Complex Ginzburg-Landau Dynamics The 2D complex Ginzburg-Landau (CGL) equation is given by ∂ ˜A(⃗r,t ) ∂t = ˜A(⃗r,t ) + (1 +iα)∇2 ˜A(⃗r,t ) − (1 +iβ)|A|2 ˜A(⃗x,t ) (S5) where ˜A(⃗r,t ) = |A(⃗r,t )| exp(iφ(⃗r,t )) is a complex field with spatiotemporal phase field φ(⃗r,t ) and amplitude |A(⃗r,t )|. Here, α and β are the linear and nonlinear dispersions ...

    work page

  7. [7]

    Construction of the Objective Function The estimator is framed as a binary classifier Cθ(⃗x,⃗x′) = σ(Sθ(⃗x,⃗x′)) where Sθ(⃗x,⃗x′) is the corre- sponding logit parametrized byθ, andσ(z) = 1/(1+e−z) is the logistic sigmoid function. This classifier is trained to receive an adjacent pair of states ( ⃗x,⃗x′) and return a prediction probability Cθ(⃗x,⃗x′)∈ (0,...

    work page

  8. [8]

    These constraints increase the specificity of the learned representations while retain- ing sufficient flexibility to capture general irreversible dy- namics

    Universality of Representations The logit S(⃗x,⃗x′) is constructed to capture the log ratio between forward and reverse joint transition proba- bilities, with suitable constraints imposed on the learned matrices Aθ and Bθ. These constraints increase the specificity of the learned representations while retain- ing sufficient flexibility to capture general ...

    work page

Show all 63 references
  1. [9]

    First fix ϵ >0, then consider N = 1 and choose the monomial basis representation ϕ(⃗x) = 1 x x2 ··· xda ⊺ (S42) soM =da + 1

    (S41) 19 Proof of Lemma 2. First fix ϵ >0, then consider N = 1 and choose the monomial basis representation ϕ(⃗x) = 1 x x2 ··· xda ⊺ (S42) soM =da + 1. By the Stone-Weierstrass theorem, there exists ada-degree polynomial for sufficiently highda that approximatesf(⃗x,⃗y) as sup...

  2. [10]

    Now, the antisymmetry of f(⃗x,⃗y) implies that daX i,j=0 aijyixj =− daX i,j=0 aijxiyj (S44) which leads to the result aji =−aij

    (S43) We then form the matrix A whose (i,j )-th element is given byaij, which enables us to write the above sum as a quadratic form ⃗ϕ⊺(⃗y)A⃗ϕ(⃗x). Now, the antisymmetry of f(⃗x,⃗y) implies that daX i,j=0 aijyixj =− daX i,j=0 aijxiyj (S44) which leads to the result aji =−aij. ...

  3. [11]

    With the dynamics given by Eq

    Principal Component Analysis (PCA) Estimation of Entropy Production In the linear setting, we may wish to decompose the irreversible dynamics of the system to a set of reduced co- ordinates which exhibit the largest variance by perform- ing principal component analysis (PCA) o...

  4. [12]

    Battle, C

    C. Battle, C. P. Broedersz, N. Fakhri, V. F. Geyer, J. Howard, C. F. Schmidt, and F. C. MacKintosh, Broken detailed balance at mesoscopic scales in active biological systems, Science 352, 604 (2016)

    2016

  5. [13]

    Gladrow, C

    J. Gladrow, C. P. Broedersz, and C. F. Schmidt, Nonequi- librium dynamics of probe filaments in actin-myosin net- works, Physical Review E 96, 022408 (2017)

    2017

  6. [14]

    C. W. Lynn, C. M. Holmes, W. Bialek, and D. J. Schwab, Emergence of local irreversibility in complex interacting systems, Physical Review E 106, 034102 (2022)

    2022

  7. [15]

    Seifert, Stochastic thermodynamics, fluctuation the- orems and molecular machines, Reports on Progress in Physics 75, 126001 (2012)

    U. Seifert, Stochastic thermodynamics, fluctuation the- orems and molecular machines, Reports on Progress in Physics 75, 126001 (2012)

    2012

  8. [16]

    H. Qian, S. Kjelstrup, A. B. Kolomeisky, and D. Bedeaux, Entropy production in mesoscopic stochastic thermody- namics: nonequilibrium kinetic cycles driven by chemical 26 potentials, temperatures, and mechanical forces, Journal of Physics: Condensed Matter 28, 153004 (2016)

    2016

  9. [17]

    F. S. Gnesotto, F. Mura, J. Gladrow, and C. P. Broed- ersz, Broken detailed balance and non-equilibrium dy- namics in living systems: a review, Reports on Progress in Physics 81, 066601 (2018)

    2018

  10. [18]

    F. Mura, G. Gradziuk, and C. P. Broedersz, Nonequilib- rium Scaling Behavior in Driven Soft Biological Assem- blies, Physical Review Letters 121, 038002 (2018)

    2018

  11. [19]

    T. H. Tan, G. A. Watson, Y.-C. Chao, J. Li, T. R. Gin- grich, J. M. Horowitz, and N. Fakhri, Scale-dependent irreversibility in living matter (2021)

    2021

  12. [20]

    Q. Wang, S. Kulkarni, and S. Verdu, Divergence estima- tion of continuous distributions based on data-dependent partitions, IEEE Transactions on Information Theory51, 3064 (2005)

    2005

  13. [21]

    Ziv and N

    J. Ziv and N. Merhav, A measure of relative entropy be- tween individual sequences with application to universal classification, IEEE Transactions on Information Theory 39, 1270 (1993)

    1993

  14. [22]

    Rold´ an and J

    R. Rold´ an and J. M. R. Parrondo, Entropy produc- tion and Kullback-Leibler divergence between stationary trajectories of discrete systems, Physical Review E 85, 031129 (2012)

    2012

  15. [23]

    A. C. Barato and U. Seifert, Thermodynamic Uncer- tainty Relation for Biomolecular Processes, Physical Re- view Letters 114, 158101 (2015)

    2015

  16. [24]

    T. R. Gingrich, J. M. Horowitz, N. Perunov, and J. L. England, Dissipation bounds all steady-state cur- rent fluctuations, Physical Review Letters 116, 120601 (2016)

    2016

  17. [25]

    J. M. Horowitz and T. R. Gingrich, Thermodynamic uncertainty relations constrain non-equilibrium fluctua- tions, Nature Physics 16, 15 (2020)

    2020

  18. [26]

    J. Li, J. M. Horowitz, T. R. Gingrich, and N. Fakhri, Quantifying dissipation using fluctuating currents, Na- ture Communications 10, 1666 (2019)

    2019

  19. [27]

    Rold´ an, J

    E. Rold´ an, J. Barral, P. Martin, J. M. R. Parrondo, and F. J¨ ulicher, Quantifying entropy production in active fluctuations of the hair-cell bundle from time irreversibil- ity and uncertainty relations, New Journal of Physics23, 083013 (2021)

    2021

  20. [28]

    F. S. Gnesotto, G. Gradziuk, P. Ronceray, and C. P. Broedersz, Learning the non-equilibrium dynamics of Brownian movies, Nature Communications 11, 5378 (2020)

    2020

  21. [29]

    D.-K. Kim, Y. Bae, S. Lee, and H. Jeong, Learning En- tropy Production via Neural Networks, Physical Review Letters 125, 140604 (2020)

    2020

  22. [30]

    Bae, D.-K

    Y. Bae, D.-K. Kim, and H. Jeong, Inferring dissipation maps from videos using convolutional neural networks, Physical Review Research 4, 033094 (2022)

    2022

  23. [31]

    J. Lyu, K. J. Ray, and J. P. Crutchfield, Learning en- tropy production from underdamped Langevin trajecto- ries, Physical Review E 110, 064151 (2024)

    2024

  24. [32]

    Vodret, C

    M. Vodret, C. Pacini, and C. Bongiorno, Functional de- composition and estimation of irreversibility in time se- ries via machine learning, Physical Review E110, 064310 (2024)

    2024

  25. [33]

    N. M. Boffi and E. Vanden-Eijnden, Model-free learning of probability flows: Elucidating the nonequilibrium dy- namics of flocking (2024)

    2024

  26. [34]

    N. M. Boffi and E. Vanden-Eijnden, Deep learning proba- bility flows and entropy production rates in active matter, Proceedings of the National Academy of Sciences 121, e2318106121 (2024)

    2024

  27. [35]

    Ma and M

    Z. Ma and M. Collins, Noise Contrastive Estimation and Negative Sampling for Conditional Models: Consistency and Statistical Efficiency (2018)

    2018

  28. [36]

    Kawai, J

    R. Kawai, J. M. R. Parrondo, and C. V. den Broeck, Dis- sipation: The Phase-Space Perspective, Physical Review Letters 98, 080602 (2007)

    2007

  29. [37]

    Rieder, J

    Z. Rieder, J. L. Lebowitz, and E. Lieb, Properties of a harmonic crystal in a stationary nonequilib- rium state, Journal of Mathematical Physics 8, 1073 (1967), https://pubs.aip.org/aip/jmp/article- pdf/8/5/1073/19085581/1073 1 online.pdf

    1967

  30. [38]

    Bonetto, J

    F. Bonetto, J. L. Lebowitz, and J. Lukkarinen, Fourier’s Law for a Harmonic Crystal with Self-Consistent Stochastic Reservoirs, Journal of Statistical Physics 116, 783 (2004)

    2004

  31. [39]

    Yildiz, J

    A. Yildiz, J. N. Forkey, S. A. McKinney, T. Ha, Y. E. Goldman, and P. R. Selvin, Myosin v walks hand-over-hand: Single fluorophore imaging with 1.5-nm localization, Science 300, 2061 (2003), https://www.science.org/doi/pdf/10.1126/science.1084398

    2003 doi

  32. [40]

    Yildiz, M

    A. Yildiz, M. Tomishige, R. D. Vale, and P. R. Selvin, Kinesin walks hand-over-hand, Science 303, 676 (2004), https://www.science.org/doi/pdf/10.1126/science.1093753

    2004 doi

  33. [41]

    Risken, The Fokker-Planck Equation: Methods of So- lution and Applications , edited by H

    H. Risken, The Fokker-Planck Equation: Methods of So- lution and Applications , edited by H. Haken, Springer Series in Synergetics, Vol. 18 (Springer, Berlin, Heidel- berg, 1996)

    1996

  34. [42]

    Chat´ e and P

    H. Chat´ e and P. Manneville, Phase Diagram of the Two- Dimensional Complex Ginzburg-Landau Equation, Phys- ica A: Statistical Mechanics and its Applications224, 348 (1996), arXiv:1608.07519 [nlin]

    1996 arXiv

  35. [43]

    I. S. Aranson and L. Kramer, The world of the complex Ginzburg-Landau equation, Reviews of Modern Physics 74, 99 (2002), publisher: American Physical Society

    2002

  36. [44]

    T. H. Tan, J. Liu, P. W. Miller, M. Tekant, J. Dunkel, and N. Fakhri, Topological turbulence in the membrane of a living cell, Nature Physics 16, 657 (2020)

    2020

  37. [45]

    Li, C.-W

    J. Li, C.-W. J. Liu, M. Szurek, and N. Fakhri, Measuring Irreversibility from Learned Representations of Biological Patterns, PRX Life 2, 033013 (2024)

    2024

  38. [46]

    See Supplementary Movies M1 and M2 at [URL will be inserted by publisher]

  39. [47]

    Watter, J

    M. Watter, J. Springenberg, J. Boedecker, and M. Ried- miller, Embed to control: A locally linear latent dynam- ics model for control from raw images, in Advances in Neural Information Processing Systems , Vol. 28, edited by C. Cortes, N. Lawrence, D. Lee, M. Sugiyama, and R. ...

    2015

  40. [48]

    Code Repository for LENS – https://github.com/ cj7280/LENS

  41. [49]

    K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learn- ing for image recognition, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2016)

    2016

  42. [50]

    J. L. Ba, J. R. Kiros, and G. E. Hinton, Layer normal- ization (2016), arXiv:1607.06450 [stat.ML]

    2016 arXiv

  43. [51]

    Hendrycks and K

    D. Hendrycks and K. Gimpel, Gaussian error linear units (gelus), arXiv preprint arXiv:1606.08415 (2016)

    2016 arXiv

  44. [52]

    R. Liu, J. Lehman, P. Molino, F. P. Such, E. Frank, A. Sergeev, and J. Yosinski, An intriguing failing of con- volutional neural networks and the coordconv solution (2018), arXiv:1807.03247 [cs.CV]. 27

    2018 arXiv

  45. [53]

    Biewald, Experiment tracking with weights and biases (2020), software available from wandb.com

    L. Biewald, Experiment tracking with weights and biases (2020), software available from wandb.com

    2020

  46. [54]

    Bortkiewicz, W

    M. Bortkiewicz, W. Pa lucki, V. Myers, T. Dziarmaga, T. Arczewski, L. Kuci´ nski, and B. Eysenbach, Accel- erating Goal-Conditioned RL Algorithms and Research, in International Conference on Learning Representations (2025)

    2025

  47. [55]

    J. Lyu, K. J. Ray, and J. P. Crutchfield, Learning Stochastic Thermodynamics Directly from Correlation and Trajectory-Fluctuation Currents (2025)

    2025

  48. [56]

    Stamhuis and W

    E. Stamhuis and W. Thielicke, Pivlab–towards user- friendly, affordable and accurate digital particle image velocimetry in matlab, Journal of open research software 2, 30 (2014)

    2014

  49. [57]

    Matsumoto, S.-i

    K. Matsumoto, S.-i. Sasa, and A. Dechant, Learning rate matrix and information-thermodynamic trade-off rela- tion (2025)

    2025

  50. [58]

    Mezzadri, How to generate random matrices from the classical compact groups (2007), arXiv:math-ph/0609050 [math-ph]

    F. Mezzadri, How to generate random matrices from the classical compact groups (2007), arXiv:math-ph/0609050 [math-ph]

    2007 arXiv

  51. [59]

    Sitzmann, J

    V. Sitzmann, J. Martel, A. Bergman, D. Lindell, and G. Wetzstein, Implicit neural representations with peri- odic activation functions, Advances in neural information processing systems 33, 7462 (2020)

    2020

  52. [60]

    Cox and P

    S. Cox and P. Matthews, Exponential time differencing for stiff systems, Journal of Computational Physics 176, 430 (2002)

    2002

  53. [61]

    Huber, P

    G. Huber, P. Alstrøm, and T. Bohr, Nucleation and tran- sients at the onset of vortex turbulence, Phys. Rev. Lett. 69, 2380 (1992)

    1992

  54. [62]

    van Saarloos and P

    W. van Saarloos and P. Hohenberg, Fronts, pulses, sources and sinks in generalized complex Ginzburg- Landau equations, Physica D: Nonlinear Phenomena 56, 303–367 (1992)

    1992

  55. [63]

    Nowozin, B

    S. Nowozin, B. Cseke, and R. Tomioka, f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization, in Advances in Neural Information Pro- cessing Systems, Vol. 29 (Curran Associates, Inc., 2016)

    2016

Pith tools

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