Pith. sign in

REVIEW 2 major objections 5 minor 42 references

anyakrakusuma: A Python Library for Entropic Schr\"odinger Bridges on Idealized Geometries

T0 review · 2 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read A log-domain Sinkhorn–Knopp iteration solves discrete entropic Schrödinger bridges at cost-to-regularization ratios up to 400, where the Gibbs kernel underflows double precision, yet it still reaches a marginal residual of 10⁻⁹ and recovers

desk verdict A useful, unusually candid software paper with one wrong headline number: the Gibbs kernel does not underflow double precision at their stated ratios, so the 'log-domain necessary' claim is unsupported. read the letter →

arxiv 2607.18184 v1 pith:3OONJODR submitted 2026-07-20 physics.comp-ph cs.MScs.NAmath.NAmath.OC

Sandy Hardian Susanto Herho , Dasapta Erwin Irawan , Agus Wahyu Jatmiko , Sito Fossy Biosa , Candrasa Surya Dharma , Edi Riawan , Astyka Pamumpuni , Rendy Dwi Kartiko
show 2 more authors
Rusmawan Suwarman Deny Juanda Puradimaja
This is my paper · ORCID
classification physics.comp-phcs.MScs.NAmath.NAmath.OC
keywords entropicoptimaltransportSchrödingerbridgeproblemSinkhorn–Knoppalgorithmlog-domainstabilizationlog-sum-exppoint-cloudcouplingcovarianceprincipal-axisanalysisdifferentialentropyestimation
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

This paper presents an open-source Python library that solves the discrete static Schrödinger bridge problem—the entropically regularized optimal transport coupling between two empirical point clouds—and reconstructs the intermediate-time bridge. Its central claim is that working in the log-domain of the Sinkhorn potentials is necessary, not merely convenient, at the small-regularization settings tested: costs exceed regularization by up to 400-fold, so the raw Gibbs kernel underflows double precision across most entries, yet the iteration reaches a marginal residual of 10⁻⁹ with unit marginal fidelity in every case. The strongest quantitative check is a covariance-based recovery of an imposed 90-degree reorientation in a two-moons case to within 0.07°, about forty times smaller than the estimator's uncertainty, even across a near-isotropic interval where the principal axis is unobservable. The paper also shows that endpoint differential-entropy estimates on noiseless curve-supported clouds measure curve-length ratios rather than entropy, and it carefully delineates the regimes in which each diagnostic is well defined.

What carries the argument

The load-bearing mechanism is the log-domain Sinkhorn–Knopp iteration on dual potentials f and g instead of exponential scalings, with a numerically stable log-sum-exp (the max-trick) as its primitive. This replaces products of underflowing kernel entries by shift-invariant additions, which is what lets the iteration run where exp(−C_ij/ε) sits below the double-precision floor. The same log-domain identity monitors the row-marginal residual without materializing the plan, and the bridge is rebuilt by sampling from the optimal coupling and adding Brownian-bridge Gaussian noise with variance εt(1−t).

What would settle it

Run the same four cases with a plain exponential-domain Sinkhorn implementation: if it converges to a 10⁻⁹ marginal residual without underflow at cost-to-regularization ratio 400, the paper's necessity claim collapses. Conversely, rebuild the two-moons target as the exact source cloud rotated 90° about the source centroid and check whether the recovered angle moves closer to 90.0000° and the transport cost falls by the translation contribution; if not, the rotation recovery is an artifact of the generative recipe rather than a measurement.

Watch

Extended reading notes

Core claim

The paper solves the discrete Schrödinger bridge as a quadratic transport cost plus ε times the KL divergence to the product measure, using log-domain Sinkhorn–Knopp updates with a stabilized log-sum-exp. At cost-to-regularization ratios up to 400, the Gibbs kernel underflows double precision over most entries, so exponential-domain scaling cannot run; the log-domain potentials stay bounded and reach a 10⁻⁹ marginal residual with unit fidelity in every case. Bridge reconstruction samples Brownian bridges conditioned on the optimal coupling, and diagnostics span coupling entropy, barycentric projections, kernel-density region counts, k-nearest-neighbor entropy estimates, and covariance ellips

Load-bearing premise

The 90-degree rotation recovery in the two-moons case is the paper's strongest quantitative validation, and it rests on the premise that the target is the source shape rotated exactly 90° about a common center; in the implementation the target is a fresh independent draw rotated about its own empirical centroid, so the imposed rotation holds only in expectation and the two clouds also differ by a translation of magnitude 0.56178.

Editorial extensions

If this is right

  • At cost-to-regularization ratios up to 400, entropic transport between point clouds remains solvable to 10⁻⁹ residuals, a regime exponential-domain scaling cannot enter; this extends practical entropic optimal transport to much smaller regularization than naive implementations allow.
  • The two-moons covariance test shows that a principal-axis rotation estimator can recover an imposed 90° rotation to within 0.07° even across a near-isotropic masked interval, provided the doubled angle is thresholded and unwrapped correctly.
  • On noiseless curve-supported endpoints, differences of k-nearest-neighbor entropy estimates at equal sample size estimate the ratio of curve lengths, not entropy production; this gives a concrete diagnostic check and suggests perturbing the generators to place all cases on a common footing.
  • The circle-to-circle case converges in a single sweep because its circulant cost matrix admits constant potentials; the residual at machine epsilon therefore serves as a correctness check for the solver rather than a convergence baseline.
  • Observed per-iteration contraction factors are local near-fixed-point rates; the worst-case contraction bound is vacuous at these settings, so the measured rates do not contradict classical Sinkhorn theory.
  • The factor-of-two offset in row perplexity caused by strided plan storage means reported conditional entropies understate full-plan values by log 2, and the full plan should be archived when storage permits.

Reading between the lines

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

  • If log-domain stabilization is what enables the runtimes shown here, the same potentials-based recursion should transfer directly to unbalanced and multi-marginal extensions, where dense cost-matrix storage rather than underflow becomes the binding constraint.
  • The contraction factors do not order with ε across the three fitted cases; a single-geometry ε-sweep would separate the influence of regularization from transport geometry on convergence speed, an experiment the paper leaves to future work.
  • Re-running the two-moons case with the target built by rotating the exact source cloud about the exact source centroid should move the recovered angle closer to exactly 90° and lower the transport cost by the translation contribution, offering a sharper validation of the rotation-recovery pipeline.
  • The paper's endpoint-entropy diagnosis suggests a simple testable extension: adding small isotropic noise to the noiseless curve-supported generators should make the endpoint entropy estimates well-defined and collapse the apparent net entropy production toward the two-dimensional expectation.
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

2 major / 5 minor

Summary. The paper presents anyakrakusuma, an open-source Python library for solving the discrete static Schrödinger bridge problem via a log-domain Sinkhorn–Knopp iteration, together with a diagnostic pipeline that analyzes the optimal coupling, entropic interpolation, and intermediate-time marginals. The library is demonstrated on four constructed planar benchmarks: circle-to-circle dilation, spiral-to-Gaussian-mixture fragmentation, two-moons rigid reorientation, and Lissajous-to-trefoil deformation. The main claims are that the log-domain formulation is necessary at the studied regularization parameters because the Gibbs kernel underflows double precision over most of its range, that the iteration reaches a marginal residual below 10^-9 with geometric contraction factors between 0.966 and 0.976, and that covariance-based diagnostics recover an imposed 90-degree rotation to within 0.07 degrees. The paper is unusually explicit about its limitations, including the translation confound in Case 3, temporally interpolated KDE snapshots, the factor-of-two effect of plan subsampling, and the invalidity of endpoint entropy estimates for noiseless curve-supported clouds.

Significance. If the claims are appropriately qualified, the paper is a useful software contribution. The mathematical derivation of the discrete Schrödinger bridge is standard and correctly presented, the implementation details are transparent, and the release of code, configuration files, diagnostic scripts, and NetCDF archives is a genuine strength. Several internal consistency checks are well designed, for example the recovery of log 4 from the difference of curve-supported endpoint entropy estimates and the log 4 agreement of the subsampled block entropy. The paper also carefully distinguishes descriptive fits from predictive claims and states the regimes in which each diagnostic is valid. The main scientific value lies in the reproducible solver-plus-diagnostics workflow rather than in new mathematical results. However, two load-bearing claims need correction: the numerical premise behind the 'log-domain is necessary' claim is quantitatively wrong at the reported parameters, and the Case 3 rotation-recovery claim is stronger than the generative experiment actually supports.

major comments (2)
  1. [§2.3, Eq. (41); §4 first limitation; Table 4] The central claim that the log-domain formulation is 'necessary rather than merely convenient' rests on the assertion that at cost-to-regularization ratios of 252–400 the Gibbs kernel K_ij = exp(-C_ij/ε) 'underflows double precision across most of its range.' This is numerically incorrect. exp(-400) ≈ 1.9×10^-174 and exp(-252) ≈ 1.5×10^-110, both far above the double-precision minimum around 2.2×10^-308, and even exp(-700) is still representable. True underflow would require C_ij/ε ≳ 708. Thus none of the four cases produces a kernel entry below representable range, and no exponential-domain baseline is run to demonstrate failure by another mechanism such as scaling-vector overflow. The favorable residual histories show that this implementation is stable, not that the log-domain formulation is necessary at these parameters. This claim appears in the abstract, the Discussion, and the Conc
  2. [Abstract; §4; §5] The Case 3 rotation validation is weaker than the text claims. The target is not obtained by rotating the source cloud; it is an independent fresh two-moons draw rotated by 90 degrees about its own empirical centroid. Consequently, the source and target covariance principal axes differ by 90 degrees only in expectation, not as a property of the realized clouds. The reported 89.9304° recovery with an uncertainty of ±2.8204° is computed from within-realization subsampling of a single pair of clouds and does not include the between-realization variability of the two independent draws. The phrase 'recovers the imposed ninety-degree reorientation' therefore overstates the validation: it is a consistency check of the estimator against a generative recipe. A concrete fix is to rotate the source cloud itself about its common centroid, or to generate multiple independent source/target pairs and r
minor comments (5)
  1. [Table 2 and §4] The factor-of-two bias in the conditional perplexities is acknowledged in the Discussion, but Table 2 reports perplexities in 'retained target points' without flagging that the full-plan values are approximately twice as large. Since the table is likely to be read in isolation, it should either report full-plan values or include an explicit footnote in the table itself.
  2. [Eq. (55)] The reference entropy H_diff(t) is defined for t in (0,1); the formula has a logarithmic singularity at t=0 and t=1. The text is clear in context, but adding t∈(0,1) to the equation or its surrounding sentence would prevent confusion.
  3. [§2.2] Minor typesetting: 'LL VM' should be 'LLVM'. Also, the statement that fastmath=True permits associative reordering is fine, but the claim that this is inconsequential because the updates are shift-invariant could be stated more precisely: shift invariance does not by itself guarantee identical fixed points under reordered floating-point arithmetic, although the observed residuals make this a non-issue here.
  4. [§4] The Authors state that animation rendering accounted for 82.5%–91.8% of total wall-clock time and that the solve is a minor fraction; the single-sweep Case 1 solve time includes JIT compilation. This is clearly explained, but the timings table might benefit from a column excluding one-time JIT cost for reproducibility comparisons.
  5. [Abstract] The phrase 'unit marginal fidelity' is used without definition. It is presumably the total plan mass / marginal residual check in Eq. (36); defining it in the abstract is unnecessary, but the term should be introduced exactly where Eq. (36) is discussed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's validations are against constructed ground truths or independently derived identities, and no fitted parameter is presented as a prediction.

full rationale

The paper's central derivation is the standard log-domain Sinkhorn recursion for entropic optimal transport, Eqs. (22)–(25), which is an independent computational formulation rather than a restatement of the paper's own outputs. The reported validations are comparisons against known or constructed values: the circle-to-circle case converges in one sweep because of the circulant cost structure and is checked against the closed-form solution; the Case 1 entropy difference equals log 4 through the independently derived curve-length ratio 2 log(L1/L0); the Case 3 rotation recovery of 89.9304° is measured against the generative rotation of a fresh two-moons cloud, with the paper itself noting the unintended translational component; and the block-entropy log 4 offset follows from the strided storage identity. The contraction factors are descriptive fits to the same residual histories and are not used as predictions of independent data, so they are self-referential only in a weak, non-circular sense. The paper's central 'log-domain necessary' claim rests on an asserted underflow premise that is not mathematically correct at the stated ratios (exp(-400) is representable in double precision), but that is a correctness/support issue, not a circularity: no quantity is defined in terms of another result or fitted to force a conclusion. Self-citations appear only as motivational examples of idealized solvers and are not load-bearing for the Schrödinger-bridge derivation or the numerical validations.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central computation relies on the standard entropic-OT / Schrödinger-bridge equivalence, the Gaussian Brownian-bridge marginal, and classical Sinkhorn convergence theory; these are imported from cited literature rather than derived here. The free parameters are experimental settings (ε, n, thresholds, k, subsampling, tolerance), all chosen by hand; none is fitted to make a prediction come out. No new physical entities are introduced.

free parameters (6)
  • regularization/diffusivity ε per case = 0.02, 0.05, 0.03, 0.04
    Chosen by hand for each of the four demonstrations; controls the sharpness of the coupling and enters the bridge variance εt(1−t). The reported contraction rates, entropies, and recovered rotation all depend on it.
  • point-cloud size n=m = 1000
    Fixed for all cases; dimensions the residual histories, KL estimator finite-sample behavior, and plan-storage subsampling. Not swept in this study.
  • eccentricity resolution threshold = e ≥ 0.25
    Frames below this threshold are masked from the principal-axis analysis; the rotation-recovery number depends on which frames are considered resolved.
  • KL estimator k = 5
    Number of nearest neighbors in the Kozachenko–Leonenko entropy estimator; entropy values and their uncertainty bands depend on this choice.
  • subsampling parameters = B=80, m=0.8n
    Uncertainty bands for entropy, dispersion, eccentricity, and orientation are computed from subsamples; the reported half-widths depend on these settings.
  • Sinkhorn tolerance / iteration cap = η < 1e-9; ≤ 2000 sweeps
    Defines what counts as convergence and 'marginal fidelity 1.00000000'; the residual histories and contraction factors are measured against this stopping rule.
assumptions (5)
  • standard math The Schrödinger bridge problem (2) reduces to the entropic OT coupling problem (10)/(14) via the KL chain-rule decomposition (8)–(9).
    Invoked in §2.1; this equivalence is proved in the cited literature (Léonard 2014; Chen–Georgiou–Pavon 2016) and is not re-derived.
  • domain assumption The Brownian bridge conditional marginal is Gaussian: X_t | (X_0=x, X_1=y) ~ N((1−t)x + ty, εt(1−t)I_d) (eq. 19).
    Used in §2.2 for the bridge sampler and in §2.4 for the entropy reference (55). Assumes the reference dynamics is pure Brownian diffusion (1).
  • standard math For positive marginals, Sinkhorn scalings exist and are unique, and the iteration converges geometrically in the Hilbert projective metric (refs. [39,40,23]).
    Basis for expecting geometric residual decay in §2.2 and §2.4. The paper explicitly notes the worst-case Birkhoff rate is vacuous at these ε values.
  • domain assumption The Kozachenko–Leonenko estimator (53) is a consistent entropy estimator for absolutely continuous densities on R^d.
    Invoked in §2.4. The paper itself flags that endpoint clouds of cases 1 and 4 lie on Lebesgue-null curves and that those estimates are not true entropies.
  • domain assumption Subsampling standard deviations obey the √n rescaling (60) and 95% bands are Gaussian.
    Used to attach uncertainties to all frame-wise descriptors in §2.4; depends on the statistics converging at n^{-1/2}.

how reviews work

0 comments
Cite this review

Pith. "Pith review of anyakrakusuma: A Python Library for Entropic Schr\"odinger Bridges on Idealized Geometries." pith.science (2026). https://pith.science/paper/3OONJODR

@misc{pith2026260718184,
  author       = {Pith},
  title        = {Pith review of: anyakrakusuma: A Python Library for Entropic Schr\"odinger Bridges on Idealized Geometries},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3OONJODR}},
  note         = {Machine review of arXiv:2607.18184}
}
abstract

We present anyakrakusuma, an open-source Python library that solves the discrete static Schr\"{o}dinger bridge problem, the entropically regularized counterpart of optimal transport, through a log-domain Sinkhorn--Knopp iteration and reconstructs the entropic interpolation between two empirical point clouds. The solver is paired with a diagnostic pipeline that characterizes the optimal coupling and the intermediate distributions through information-theoretic and geometric measures. We exercise the library on four idealized planar cases spanning a circle-to-circle dilation, a spiral-to-mixture fragmentation, a rigid reorientation of two moons, and a Lissajous-to-trefoil deformation. The log-domain formulation is necessary rather than merely convenient at the parameters studied, where the cost-to-regularization ratio reaches four hundred and the Gibbs kernel underflows double precision across most of its range; the iteration nonetheless attains a marginal residual of $10^{-9}$ and unit marginal fidelity in every case. Residual histories decay geometrically over approximately eight decades at per-iteration contraction factors between $0.966$ and $0.976$, which are local rates near the fixed point that lie many orders of magnitude below the worst-case Hilbert-metric bound. The covariance analysis recovers an imposed ninety-degree reorientation to within $0.07^\circ$, roughly forty times smaller than its uncertainty, across a masked interval of near-isotropy on which the principal axis is unobservable. The diagnostics are reported with explicit attention to the regimes in which each is well defined, including the differential entropy, which is meaningful only on the open interpolation interval. The presented cases are constructed rather than measured; quantitative application to empirical point clouds requires further study.

Figures

Figures reproduced from arXiv: 2607.18184 by the authors.

Figure 1
Figure 1. Convergence of the log-domain Sinkhorn iteration for the three cases that produced a fittable residual [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. shows the barycentric map of each case as displacement arrows from a strided subset of ninety source points to their conditional-mean images, overlaid on the source and target clouds. The arrows in figure 2(a) are directed radially outward and are of visually uniform length. Those in figure 2(b) converge onto the four mixture centers, the arrow bundles originating along the turns of the spiral and terminating in tig… view at source ↗
Figure 3
Figure 3. Gaussian kernel density estimates of the bridge marginal [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Frame-wise descriptors of the bridge marginal against interpolation time [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 4 canonical work pages

  1. [1]

    Iterative Bregman Projections for Regularized Transportation Problems.SIAM J

    Benamou, J.-D.; Carlier, G.; Cuturi, M.; Nenna, L.; Peyr´ e, G. Iterative Bregman Projections for Regularized Transportation Problems.SIAM J. Sci. Comput.2015,37(2), A1111–A1138.https://doi.org/10.1137/ 141000439

    2015

  2. [2]

    Efficient multivariate entropy estimation viak-nearest neighbour distances.Ann

    Berrett, T.B.; Samworth, R.J.; Yuan, M. Efficient multivariate entropy estimation viak-nearest neighbour distances.Ann. Stat.2016,47(1), 288–318.https://doi.org/10.1214/18-AOS1688

    doi:10.1214/18-aos1688 2016

  3. [3]

    Polar factorization and monotone rearrangement of vector-valued functions.Commun

    Brenier, Y. Polar factorization and monotone rearrangement of vector-valued functions.Commun. Pure Appl. Math.1991,44(4), 375–417.https://doi.org/10.1002/cpa.3160440402

    doi:10.1002/cpa.3160440402 1991

  4. [4]

    Convergence of Entropic Schemes for Optimal Transport and Gradient Flows.SIAM J

    Carlier, G.; Duval, V.; Peyr´ e, G.; Schmitzer, B. Convergence of Entropic Schemes for Optimal Transport and Gradient Flows.SIAM J. Math. Anal.2017,49(2), 1385–1418.https://doi.org/10.1137/15M1050264

    work page doi:10.1137/15m1050264 2017

  5. [5]

    On the Relation Between Optimal Transport and Schr¨ odinger Bridges: A Stochastic Control Viewpoint.J

    Chen, Y.; Georgiou, T.T.; Pavon, M. On the Relation Between Optimal Transport and Schr¨ odinger Bridges: A Stochastic Control Viewpoint.J. Optim. Theory Appl.2016,169, 671–691.https://doi.org/10.1007/ s10957-015-0803-z

    2016

  6. [6]

    Scaling algorithms for unbalanced optimal transport problems.Math

    Chizat, L.; Peyr´ e, G.; Schmitzer, B.; Vialard, F.-X. Scaling algorithms for unbalanced optimal transport problems.Math. Comp.2018,87, 2563–2609.https://doi.org/10.1090/mcom/3303

    doi:10.1090/mcom/3303 2018

  7. [7]

    Sinkhorn Distances: Lightspeed Computation of Optimal Transport

    Cuturi, M. Sinkhorn Distances: Lightspeed Computation of Optimal Transport. InAdvances in Neural Information Processing Systems 26 (NeurIPS 2013); Curran Associates, Inc., 2013; pp. 2292–2300.https: //proceedings.neurips.cc/paper/2013/hash/af21d0c97db2e27e13572cbf59eb343d-Abstract.html

    2013

  8. [8]

    Diffusion Schr¨ odinger Bridge with Applications to Score- Based Generative Modeling

    De Bortoli, V.; Thornton, J.; Heng, J.; Doucet, A. Diffusion Schr¨ odinger Bridge with Applications to Score- Based Generative Modeling. InAdvances in Neural Information Processing Systems 34 (NeurIPS 2021); Curran Associates, Inc., 2021; pp. 17695–17709.https://proceedings.neurips.cc/paper/2021/hash/ 940392f5f32a7ade1cc201767cf83e31-Abstract.html

    2021

Show all 42 references
  1. [9]

    On the Kozachenko–Leonenko entropy estimator.J

    Delattre, S.; Fournier, N. On the Kozachenko–Leonenko entropy estimator.J. Stat. Plan. Inference2017, 185, 69–93.https://doi.org/10.1016/j.jspi.2017.01.004

    2017 doi

  2. [10]

    https://doi.org/10.1201/9780429246593

    Efron, B.; Tibshirani, R.J.An Introduction to the Bootstrap; Chapman & Hall/CRC: New York, NY, 1994. https://doi.org/10.1201/9780429246593

    1994 doi

  3. [11]

    Random fields and diffusion processes

    F¨ ollmer, H. Random fields and diffusion processes. In ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour XV–XVII, 1985–87; Hennequin, P.L., Ed.; Lecture Notes in Mathematics, Vol. 1362; Springer: Berlin, Heidelberg, Germany, 1988; pp. 101–203.https://doi.org/10.1007/BFb0086180

    1985 doi

  4. [12]

    On the scaling of multidimensional matrices.Linear Algebra Appl.1989,114–115, 717–735.https://doi.org/10.1016/0024-3795(89)90490-4

    Franklin, J.; Lorenz, J. On the scaling of multidimensional matrices.Linear Algebra Appl.1989,114–115, 717–735.https://doi.org/10.1016/0024-3795(89)90490-4

    1989 doi

  5. [13]

    What every computer scientist should know about floating-point arithmetic.ACM Comput

    Goldberg, D. What every computer scientist should know about floating-point arithmetic.ACM Comput. Surv.1991,23(1), 5–48.https://doi.org/10.1145/103162.103163

    1991

  6. [14]

    Array programming with NumPy.Nature2020,585, 357–362

    Harris, C.R.; Millman, K.J.; van der Walt, S.J.; Gommers, R.; Virtanen, P.; Cournapeau, D.; Wieser, E.; Taylor, J.; Berg, S.; Smith, N.J.; Kern, R.; Picus, M.; Hoyer, S.; van Kerkwijk, M.H.; Brett, M.; Haldane, A.; del R ´ ıo, J.F.; Wiebe, M.; Peterson, P.; G´ erard-Marchant, ...

    doi

  7. [15]

    Herho, S.H.S.; Anwar, I.P.; Khadami, F.; Handayani, A.P.; Sujatmiko, K.A.; Kasim, K.; Suwarman, R.; Irawan, D.E.dewi-Kadita: a Python library for idealized fish schooling simulation with entropy-based diagnostics.J. Phys. Commun.2026,10(6), 065002.https://doi.org/10.1088/2399-...

    2026 doi

  8. [16]

    wave-attenuation-1d: An Idealized One-Dimensional Framework for Wave Attenuation through Coastal Vegetation using Numba-Accelerated Shallow Water Equations.J

    Herho, S.H.S.; Anwar, I.P.; Khadami, F.; Ndruru, T.R.E.B.N.; Suwarman, R.; Irawan, D.E. wave-attenuation-1d: An Idealized One-Dimensional Framework for Wave Attenuation through Coastal Vegetation using Numba-Accelerated Shallow Water Equations.J. Theor. Appl. Mech.2026,56(1), ...

    2026 doi

  9. [17]

    Reappraising double pendulum dynamics across multiple computational platforms.CLEI Electron

    Herho, S.H.S.; Fajary, F.R.; Herho, K.E.P.; Anwar, I.P.; Suwarman, R.; Irawan, D.E. Reappraising double pendulum dynamics across multiple computational platforms.CLEI Electron. J.2025,28(1), 10.https: //doi.org/10.19153/cleiej.28.1.10

    2025 doi

  10. [18]

    OptionMC: a Python package for Monte Carlo pricing of European options.Int

    Herho, S.H.S.; Kaban, S.N.; Nugraha, C. OptionMC: a Python package for Monte Carlo pricing of European options.Int. J. Data Sci.2025,6(2), 70–84.https://doi.org/10.18517/ijods.6.2.70-84.2025

    2025 doi

  11. [19]

    kh2d-solver: a Python library for idealized two-dimensional incompressible Kelvin–Helmholtz instability

    Herho, S.H.S.; Trilaksono, N.J.; Fajary, F.R.; Napitupulu, G.; Anwar, I.P.; Khadami, F.; Irawan, D.E. kh2d-solver: a Python library for idealized two-dimensional incompressible Kelvin–Helmholtz instability. Appl. Comput. Mech.2025,19(2), 125–156.https://doi.org/10.24132/acm.2025.1040

    2025

  12. [20]

    xarray: N-D Labeled Arrays and Datasets in Python.J

    Hoyer, S.; Hamman, J.J. xarray: N-D Labeled Arrays and Datasets in Python.J. Open Res. Softw.2017, 5(1), 10.https://doi.org/10.5334/jors.148

    2017 doi

  13. [21]

    Matplotlib: A 2D Graphics Environment.Comput

    Hunter, J.D. Matplotlib: A 2D Graphics Environment.Comput. Sci. Eng.2007,9(3), 90–95.https: //doi.org/10.1109/MCSE.2007.55

    2007 doi

  14. [22]

    An Open-Source Pseudo-Spectral Solver for Idealized Korteweg–de Vries Soliton Simulations.Water2026,18(7), 779.https://doi.org/10.3390/w18070779

    Irawan, D.E.; Herho, S.H.S.; Pamumpuni, A.; Kartiko, R.D.; Khadami, F.; Anwar, I.P.; Sujatmiko, K.A.; Handayani, A.P.; Fajary, F.R.; Suwarman, R. An Open-Source Pseudo-Spectral Solver for Idealized Korteweg–de Vries Soliton Simulations.Water2026,18(7), 779.https://doi.org/10.3...

    doi

  15. [23]

    The Sinkhorn–Knopp Algorithm: Convergence and Applications.SIAM J

    Knight, P.A. The Sinkhorn–Knopp Algorithm: Convergence and Applications.SIAM J. Matrix Anal. Appl. 2008,30(1), 261–275.https://doi.org/10.1137/060659624

    2008 doi

  16. [24]

    Sample Estimate of the Entropy of a Random Vector.Probl

    Kozachenko, L.F.; Leonenko, N.N. Sample Estimate of the Entropy of a Random Vector.Probl. Inf. Transm. 21 1987,23(2), 95–101. English translation of Problemy Peredachi Informatsii, 23(2), 9–16

    1987

  17. [25]

    Estimating mutual information.Phys

    Kraskov, A.; St¨ ogbauer, H.; Grassberger, P. Estimating mutual information.Phys. Rev. E2004,69(6), 066138.https://doi.org/10.1103/PhysRevE.69.066138

    doi

  18. [26]

    On Information and Sufficiency.Ann

    Kullback, S.; Leibler, R.A. On Information and Sufficiency.Ann. Math. Stat.1951,22(1), 79–86.https: //doi.org/10.1214/aoms/1177729694

    1951

  19. [27]

    Numba: A LL VM-based Python JIT compiler

    Lam, S.K.; Pitrou, A.; Seibert, S. Numba: A LL VM-based Python JIT compiler. InProceedings of the Second Workshop on the LL VM Compiler Infrastructure in HPC (LL VM-HPC 2015); ACM: New York, NY, USA, 2015; pp. 1–6.https://doi.org/10.1145/2833157.2833162

    2015

  20. [28]

    A survey of the Schr¨ odinger problem and some of its connections with optimal transport

    L´ eonard, C. A survey of the Schr¨ odinger problem and some of its connections with optimal transport. Discrete Contin. Dyn. Syst.2014,34(4), 1533–1574.https://doi.org/10.3934/dcds.2014.34.1533

    2014 doi

  21. [29]

    Mardia, K.V.; Jupp, P.E.Directional Statistics; Wiley Series in Probability and Statistics; John Wiley & Sons: Chichester, UK, 2000.https://doi.org/10.1002/9780470316979

    doi

  22. [30]

    Monge’s problem with a quadratic cost by the zero-noise limit ofh-path processes.Probab

    Mikami, T. Monge’s problem with a quadratic cost by the zero-noise limit ofh-path processes.Probab. Theory Relat. Fields2004,129, 245–260.https://doi.org/10.1007/s00440-004-0340-4

    doi

  23. [31]

    Computational Optimal Transport: With Applications to Data Science.Found

    Peyr´ e, G.; Cuturi, M. Computational Optimal Transport: With Applications to Data Science.Found. Trends Mach. Learn.2019,11(5–6), 355–607.https://doi.org/10.1561/2200000073

    2019 doi

  24. [32]

    Large Sample Confidence Regions Based on Subsamples under Minimal As- sumptions.Ann

    Politis, D.N.; Romano, J.P. Large Sample Confidence Regions Based on Subsamples under Minimal As- sumptions.Ann. Stat.1994,22(4), 2031–2050.https://doi.org/10.1214/aos/1176325770

    1994

  25. [33]

    NetCDF: an interface for scientific data access.IEEE Comput

    Rew, R.K.; Davis, G.P. NetCDF: an interface for scientific data access.IEEE Comput. Graph. Appl.1990, 10(4), 76–82.https://doi.org/10.1109/38.56302

    1990 doi

  26. [34]

    Stabilized Sparse Scaling Algorithms for Entropy Regularized Transport Problems.SIAM J

    Schmitzer, B. Stabilized Sparse Scaling Algorithms for Entropy Regularized Transport Problems.SIAM J. Sci. Comput.2019,41(3), A1443–A1481.https://doi.org/10.1137/16M1106018

    2019 doi

  27. [35]

    Schr¨ odinger, E.¨Uber die Umkehrung der Naturgesetze.Sitzungsber. Preuß. Akad. Wiss., Phys.-Math. Kl. 1931, 144–153

    1931

  28. [36]

    Sur la th´ eorie relativiste de l’´ electron et l’interpr´ etation de la m´ ecanique quantique.Ann

    Schr¨ odinger, E. Sur la th´ eorie relativiste de l’´ electron et l’interpr´ etation de la m´ ecanique quantique.Ann. Inst. Henri Poincar´ e1932,2(4), 269–310

  29. [37]

    On optimal and data-based histograms.Biometrika1979,66(3), 605–610.https://doi.org/ 10.1093/biomet/66.3.605

    Scott, D.W. On optimal and data-based histograms.Biometrika1979,66(3), 605–610.https://doi.org/ 10.1093/biomet/66.3.605

    doi

  30. [38]

    https://doi.org/10.1201/9781315140919

    Silverman, B.W.Density Estimation for Statistics and Data Analysis; Routledge: New York, NY, 1998. https://doi.org/10.1201/9781315140919

    1998 doi

  31. [39]

    A Relationship Between Arbitrary Positive Matrices and Doubly Stochastic Matrices.Ann

    Sinkhorn, R. A Relationship Between Arbitrary Positive Matrices and Doubly Stochastic Matrices.Ann. Math. Stat.1964,35(2), 876–879.https://doi.org/10.1214/aoms/1177703591

    1964

  32. [40]

    Concerning nonnegative matrices and doubly stochastic matrices.Pac

    Sinkhorn, R.; Knopp, P. Concerning nonnegative matrices and doubly stochastic matrices.Pac. J. Math. 1967,21(2), 343–348.https://doi.org/10.2140/pjm.1967.21.343

    1967 doi

  33. [41]

    338; Springer: Berlin, Heidelberg, Germany, 2009.https://doi.org/10.1007/978-3-540-71050-9

    Villani, C.Optimal Transport: Old and New; Grundlehren der mathematischen Wissenschaften, Vol. 338; Springer: Berlin, Heidelberg, Germany, 2009.https://doi.org/10.1007/978-3-540-71050-9

    2009 doi

  34. [42]

    SciPy 1.0: fundamental algorithms for scientific computing in Python.Nat

    Virtanen, P.; Gommers, R.; Oliphant, T.E.; Haberland, M.; Reddy, T.; Cournapeau, D.; Burovski, E.; Pe- terson, P.; Weckesser, W.; Bright, J.; van der Walt, S.J.; Brett, M.; Wilson, J.; Millman, K.J.; Mayorov, N.; Nelson, A.R.J.; Jones, E.; Kern, R.; Larson, E.; Carey, C.J.; Po...

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.