Pith. sign in

REVIEW 3 major objections 4 minor 29 references

Reflected Schrodinger Bridge Problem over Sub-Riemannian Manifold

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper proves that the reflected sub-Riemannian Schrödinger bridge has a unique solution under a non-characteristic boundary condition, with the optimal control given by an oblique reflection field and a forward-backward PDE system.

desk verdict Genuinely new oblique reflection construction for reflected sub-Riemannian Schrödinger bridges, but the existence of the reference process is unproved and the main theorems lean on an unpublished preprint. read the letter →

arxiv 2607.17904 v4 pith:FF2VF7CE submitted 2026-07-20 math.OC

Daniel Owusu Adu This is my paper
classification math.OC MSC 49Q2253C1760J6049K45
keywords Schrödingerbridgesub-RiemanniangeometryobliquereflectiondegeneratediffusionhypoellipticPDE-basedSinkhornnon-characteristicboundaryunderactuatedcontrol
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper solves the reflected sub-Riemannian Schrödinger bridge problem: transporting an initial probability distribution to a final one at minimum energy when the dynamics are underactuated (more ambient dimensions than control/noise directions) and hard state constraints keep the paths inside a bounded domain. It shows that normal Euclidean reflection is incompatible with the horizontal subbundle, so it constructs an intrinsic oblique reflection field — the unique horizontal direction maximally pointing inward — and proves that under Hörmander's bracket condition and a non-characteristic boundary assumption, the reflected reference process admits a smooth strictly positive transition density. This makes the Schrödinger bridge well-posed: the optimal control has the form ε g^T ∇ log φ^r, characterized by a forward-backward PDE system with asymmetric boundary conditions (oblique Neumann backward, normal no-flux forward). Because explicit transition densities are unavailable, the paper develops a PDE-based Sinkhorn iteration that enforces the boundary conditions directly. A concrete example on a solid torus demonstrates the construction, and a counterexample on a cylinder shows the non-characteristic condition is sharp.

What carries the argument

The key object is the intrinsic oblique reflection field r(x)=G(x)n(x)/||g(x)^T n(x)|| defined on the boundary: among all horizontal directions of the form g(x)a with ||a||=1, it is the one most aligned with the inward Euclidean normal n(x). When g=Id it reduces to the usual normal reflection. This field is what makes the reflected SDE well-posed and keeps the diffusion inside the horizontal subbundle; it also induces the asymmetric boundary conditions that are the signature of the paper's PDE system — an oblique Neumann condition r·∇φ=0 for the backward Schrödinger factor and a normal no-flux condition for the forward factor. The PDE-based Sinkhorn iteration alternates these two evolutions

What would settle it

On the cylinder B×S^1 with the distribution (25), at the boundary curve θ=φ±π/2 the quantity g(x)^T n(x) vanishes; if the reflected SDE with the construction (7) is simulated there and either leaves the domain or produces a non-positive density, the non-characteristic condition is confirmed necessary rather than merely sufficient.

Watch

Extended reading notes

Core claim

The central claim is that the reflected sub-Riemannian Schrödinger bridge problem is well-posed when the reflection is chosen intrinsically rather than prescribed as the Euclidean normal. The paper defines the oblique reflection field r(x)=G(x)n(x)/||g(x)^T n(x)||, the unique horizontal direction that is maximally inward-pointing, and proves that under Assumptions 2.1–2.3 the degenerate reflected diffusion has a smooth, strictly positive transition density on X×X. Consequently the Schrödinger bridge admits a unique solution, and the optimal control is u^r = ε g^T ∇ log φ^r, where φ^r is the backward heat evolution of the final density factor. The reflection mechanism preserves the horizontal

Load-bearing premise

At every boundary point the horizontal distribution must have a nonzero inward normal component (g(x)^T n(x)≠0); if it vanishes, no horizontal reflection can be uniformly inward-pointing and the claimed well-posedness collapses.

Editorial extensions

If this is right

  • For underactuated multi-agent systems in confined workspaces, the result gives a principled way to compute minimum-energy reconfiguration that respects both nonholonomic kinematics and hard walls.
  • For constrained generative modeling on manifolds with boundary, it provides a mathematically grounded alternative to normal-reflection schemes, which would push samples off the data manifold.
  • The PDE-based Sinkhorn iteration extends the Sinkhorn idea to hypoelliptic settings where transition densities are not explicit, requiring only the generator and the boundary conditions.
  • The sharp non-characteristic condition gives a design guideline: a domain/distribution pair satisfying g^T n ≠ 0 is necessary for global well-posedness.

Reading between the lines

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

  • The same oblique-reflection construction could serve as a standalone definition of reflected sub-Riemannian Brownian motion on bounded domains, independent of the Schrödinger bridge context.
  • Because the proof relies only on the smooth density and boundary behavior, the asymmetric boundary-condition structure likely persists for sticky or Wentzell boundary conditions, as the conclusion speculates; a concrete test would be to add a sticky term and re-derive the factor equations.
  • The zero-noise limit of the bridge may converge to a geodesic or Wasserstein-type problem on the sub-Riemannian domain, but this requires an independent limiting argument beyond the paper.
Share X Bluesky LinkedIn Reddit HN

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 formulates a reflected Schrödinger bridge problem for degenerate sub-Riemannian diffusions on a bounded domain. It introduces an oblique reflection field r = G n / ||g^T n||, which lies in the horizontal subbundle, and under Hörmander's condition and a non-characteristic boundary assumption claims that the reflected reference process has a smooth, strictly positive transition density and that the Schrödinger bridge admits a unique optimal control of the form u = ε g^T ∇ log φ. The paper also proposes a PDE-based Sinkhorn iteration with asymmetric boundary conditions and illustrates the construction on a solid torus, while Example 4.2 shows that the problem can fail global well-posedness when the non-characteristic condition fails.

Significance. If the claims were fully established, the paper would be a useful unification of two active lines of work: reflected Schrödinger bridges and sub-Riemannian Schrödinger bridges. The geometric choice of the reflection field, the asymmetric boundary conditions, and the counterexample to the non-characteristic condition are original and potentially valuable. The numerical example is suggestive. However, the central theoretical contributions are not proved in the manuscript: the existence of the reference measure rests on a questionable application of a classical reflected-SDE theorem, and the main optimal-control and duality results are deferred to the author's own unpublished preprint [3]. No code or machine-checked proof is supplied. The significance is therefore conditional on substantial missing mathematical support.

major comments (3)
  1. [§2, Proposition 2.4, Eq. (8)] The proof of existence of the reflected reference process (3) is not valid as written. It verifies only n^T a n ≥ ε κ^2 > 0, which controls the normal quadratic form of a = ε G = ε g g^T. Since m < d, a is rank-deficient and the strong-ellipticity hypothesis of the cited [20, Ch. IV, Thm 7.2] is not satisfied. The manuscript provides no alternative theorem or condition covering degenerate reflected SDEs under Hörmander's condition and a non-characteristic boundary. Because the measure R_{ε,μ0} defined through (3) is the reference measure for the bridge problem (4)–(6), this gap affects well-posedness of the entire construction.
  2. [§2, Theorem 2.7; §3, Theorem 3.1] These are the central results of the paper, but neither is proved. Theorem 2.7 is justified by “we refer to [3,22]” and Theorem 3.1 by “we state the following result and refer to [3]”, where [3] is the author's own unpublished preprint. The claimed new ingredients — preservation of the oblique reflection under h-transform, well-posedness of the forward-backward system with the asymmetric boundary conditions, and strong duality — are not demonstrated. A standalone journal submission must either contain the proofs or give a precise and verifiable reduction to published theorems. The current dependence on an inaccessible preprint is not acceptable for the main results.
  3. [§2, Proposition 2.6] The proof of smooth, strictly positive transition density for the reflected degenerate diffusion is too compressed to be checkable. It asserts that [9, Prop 1.8] makes each boundary point a “very good point” and that (8) suffices to satisfy [9, Thm 1.7], but it does not identify which hypotheses of [9] correspond to Assumptions 2.1–2.3 or verify the required uniform bracket bounds on the compact boundary. Positivity is solely cited to [10, Thm 5.36]. Since Proposition 2.6 is load-bearing for Theorem 2.7, the proof should be either fully written out or replaced by a precise statement of the applicable theorem with all conditions verified.
minor comments (4)
  1. [§1, Eq. (3)] The reference measure R_{ε,μ0} is introduced with the qualifier “whenever the solution exists”, but the formulation of (4) depends on R_{ε,μ0} before its existence is established. This is a presentation issue given Proposition 2.4, but the ordering should be clarified.
  2. [§2, Lemma 2.5] The martingale argument in the proof is stated circuitously: “by integrating both sides over [0,t], we get that ...” is unnecessary, and “under Assumption 2.1” should more precisely be “by the boundary condition r·∇φ = 0 on ∂X”. The conclusion is correct, but the presentation is confusing.
  3. [§2, Theorem 2.7] The notation in the final display contains typos such as “pr_{tfϵ}(x,y)”; it should read p^r_{t_f, ε}(x, y) or similar. Also, the system is displayed without punctuation, which makes the equations hard to parse.
  4. [§4, Example 4.1] The numerical section is illustrative only: no discretization scheme, mesh size, convergence criterion, or comparison with an independent solution is given for the PDE Sinkhorn iteration. This does not affect the theoretical claims, but the reader cannot verify that the computed control actually solves (16).

Circularity Check

2 steps flagged · score 5.0 of 10

Central optimal-control and duality theorems are deferred to the author's own prior preprint [3]; self-citation is load-bearing.

  1. self citation load bearing [Theorem 2.7 (Section 2, after Proposition 2.6)]
    "We state the following result and refer the reader to [3, 22]."

    The main existence/uniqueness theorem and the optimal-control formula u^r = epsilon g^T grad log phi^r are not proved in the paper. The proof is delegated to [3], a prior arXiv preprint by the present author and co-authors, and to [22], a survey of the classical (non-reflected) Schrodinger problem. The reflection-specific complications are only described in prose ('our reflection mechanism introduces two novel features in the proof'), so the central claim is not independently derived here. This is load-bearing self-citation rather than a definitional reduction.

  2. self citation load bearing [Theorem 3.1 (Section 3)]
    "We state the following result and refer to [3]."

    The strong-duality / Eulerian characterization that underlies the PDE-based Sinkhorn algorithm is deferred to the author's own prior work [3] without proof. Since this theorem is the basis for the numerical scheme and its convergence claims, the derivation chain again terminates in a self-citation rather than in a self-contained argument. No independent machine-checked, code-reproduced, or externally falsifiable support for this specific reflected sub-Riemannian statement is provided.

full rationale

Scoring: 5/10. I find no self-definitional or fitted-input circularity: the transition-density assertion (Prop. 2.6) is anchored to external works [9,10], and Prop. 2.4, whatever its correctness problems, invokes the external Ikeda-Watanabe theorem rather than a self-derived premise. The circularity that is present is the repeated delegation of the paper's central theorems to the author's own prior preprint [3]. Theorem 2.7 -- the unique solvability and the optimal control formula -- is stated with 'refer the reader to [3,22]' and only prose about the reflection-specific novelties. Theorem 3.1, the strong-duality/Eulerian formulation on which the numerical Sinkhorn scheme rests, is simply 'refer to [3]'. Because [3] is not machine-checked, independently reproduced, or shown to include the boundary conditions introduced here, these citations are not independent evidence under the stated standard. This is load-bearing self-citation rather than a definitional reduction; the paper does contain independent external anchors for parts of the chain, so the score is 5 rather than 6-8. The apparent misuse of [20]'s Theorem 7.2 for a rank-deficient diffusion (only n^T a n > 0 is verified, not uniform ellipticity) is a serious correctness gap, but it is not a circularity and is therefore not counted in the score.

Assumptions & free parameters 2 free parameters · 7 assumptions · 0 invented entities

No new particles, fields, dimensions, or entities are postulated. The oblique reflection r is constructed from the known geometry (G and n), not assumed ad hoc. The central theoretical claims rest on the geometric assumptions 2.1–2.3 and on external theorems cited from the literature; the two main optimality theorems are deferred to the author's own preprint [3], which is not independently verified here.

free parameters (2)
  • α = 0.25 (simulation)
    Coupling parameter in the horizontal distribution (17); chosen for the solid-torus example (Figure 5). It does not enter the general theorems but affects the numerical demonstration.
  • σ0, σf
    Widths of the initial and final Gaussian-mixture densities (19)–(20); chosen by hand to define the example. They are not fitted and do not affect the central theoretical claim.
assumptions (7)
  • domain assumption Assumption 2.1: each g_i ∈ C∞(X; R^d)
    Regularity needed for SDE coefficients, hypoellipticity, and boundary analysis.
  • domain assumption Assumption 2.2: Hörmander bracket-generating condition Lie(g_1,...,g_m) = R^d
    Guarantees existence of a smooth, strictly positive transition density in Prop. 2.6.
  • domain assumption Assumption 2.3: D(x) ⊄ T_x∂X for all x ∈ ∂X, equivalently g(x)^T n(x) ≠ 0
    Load-bearing: makes r in (7) well-defined and inward-pointing; also required by Prop. 2.4. Fails in Example 4.2.
  • domain assumption μ0 and μf have strictly positive densities with respect to Popp measure m
    Needed for the Schrödinger factors and endpoint coupling equations; without strict positivity the normalization may fail.
  • standard math Cattiaux's smooth-density theorem [9, Thm 1.7] and positivity theorem [10, Thm 5.36] apply to the reflected degenerate process
    Used in Prop. 2.6; the paper provides only a brief verification of 'very good points' and boundary non-degeneracy.
  • standard math Ikeda–Watanabe reflected SDE existence theorem [20, Ch.IV Thm 7.2] applies to (3)
    Used in Prop. 2.4; questionable because the diffusion is degenerate and only n^T a n > 0 is checked, not uniform ellipticity.
  • standard math Fortet/Leonard Schrödinger-bridge theory [17,22] applies in the reflected sub-Riemannian setting
    Used in Theorem 2.7; the adaptation to oblique reflection is not proved here and is deferred to [3].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reflected Schrodinger Bridge Problem over Sub-Riemannian Manifold." pith.science (2026). https://pith.science/paper/FF2VF7CE

@misc{pith2026260717904,
  author       = {Pith},
  title        = {Pith review of: Reflected Schrodinger Bridge Problem over Sub-Riemannian Manifold},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FF2VF7CE}},
  note         = {Machine review of arXiv:2607.17904}
}
read the original abstract

We formulate and solve the reflected sub-Riemannian Schr\"odinger bridge (SB) problem: minimum-energy transport of probability distributions on a bounded domain for underactuated dynamics with degenerate diffusion and hard state constraints. The main difficulties are geometric and degeneracy: Euclidean normal reflection is generally incompatible with the horizontal subbundle, since it may push the process in directions not generated by the admissible control and noise fields. Also, the reference measure may not have full support. We address these by introducing an intrinsic oblique reflection mechanism that is compatible to the sub-Riemannian structure. Under H\"ormander and non-characteristic boundary assumptions, we prove that the reflected degenerate reference process admits a smooth, strictly positive transition density. The resulting optimal control is characterized by a forward--backward PDE system with asymmetric boundary conditions: an oblique Neumann condition for the backward factor and a normal no-flux condition for the forward factor. Since explicit transition densities are generally not available in this setting, we develop a PDE-based Sinkhorn iteration that enforces these boundary conditions directly. Our examples reveal that SB depends on the topology of the domain and the geometry of the diffusion.

Figures

Figures reproduced from arXiv: 2607.17904 by the authors.

Figure 1
Figure 1. The figure above shows uncontrolled oblique reflected diffusions on the solid torus X ⊂ R 3 driven by the pure horizontal subbundle in different viewing angles. In both cases, for a fixed z0 in the z-component of X , the corresponding conditional reference measure associated to (3) is concentrated on paths is confined to the horizontal slice. (a) SDE with oblique reflection field (b) SDE with normal reflection field… view at source ↗
Figure 2
Figure 2. Figure (a) demonstrates the sub-Riemannian reflected SDE with the oblique reflection field in (3). For this g(Xt)dWt, r(Xt) ∈ D(Xt). Figure (b) plots dqt = √ ϵ dBt + n(qt)dηt. In this case, normal reflection is the geometrically natural reflection field because the diffusion acts in all ambient directions. Geometrically, Assumption (2.3) is equivalent to g(x) ⊤n(x) ̸= 0 for all x ∈ ∂X . Under Assumption (2.3) the op… view at source ↗
Figure 3
Figure 3. Iterative structure of the reflected Schrödinger system (16a)-(16b). The for￾ward evolution (top) propagates the dual factor φˆ under the no-flux condition [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Isosurfaces of the initial and final densities defined in (19) and (20), respectively. 4. Numerical Example We demonstrate our theoretical result with an example. For a non-trivial demonstration, the minimum dimension is 3 because one can genuinely have non-holonomic c…
Figure 5
Figure 5. Figure 5: The sample paths are generated by the controlled reflected dynamics (5)-(6) dictated by D in (17) with α = 0.25. The black circle markers indicate samples drawn from ρ0 in Figure 4a. The trajectories are steered by u ⋆ ϵ = ε g⊤∇ log φϵ toward the prescribed final distr…
Figure 6
Figure 6. Figure 6: This demonstrates the evolution of the optimal density in [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: The above compares the reparameterized distance in (22) of the two reflection fields, the normal field n in (21) and the oblique field r in (23) from the horizontal distribution D in (17). The normal field n has a nonzero distance from D(x) except on the equatorial bou…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 4 linked inside Pith

  1. [3]

    From schrodinger bridge to optimal transport over sub-riemannian manifolds.arXiv preprint arXiv:2605.11429, 2026

    Daniel Owusu Adu, Karthik Elamvazhuthi, and Bahman Gharesifard. From schrodinger bridge to optimal transport over sub-riemannian manifolds.arXiv preprint arXiv:2605.11429, 2026

    arXiv 2026

  2. [9]

    Time reversal of diffusion processes with a boundary condition.Stochastic processes and their Applications, 28(2):275–292, 1988

    Patrick Cattiaux. Time reversal of diffusion processes with a boundary condition.Stochastic processes and their Applications, 28(2):275–292, 1988

    1988

  3. [10]

    Stochastic calculus and degenerate boundary value problems

    Patrick Cattiaux. Stochastic calculus and degenerate boundary value problems. InAnnales de l’institut Fourier, volume 42, pages 541–624, 1992

    1992

  4. [22]

    A survey of the Schrödinger problem and some of its connections with optimal transport.Probability Surveys, 11:1–66, 2014

    Christian Léonard. A survey of the Schrödinger problem and some of its connections with optimal transport.Probability Surveys, 11:1–66, 2014

    2014

  5. [1]

    Flow matching for averaged systems.arXiv preprint arXiv:2510.14514, 2025

    Daniel Owusu Adu and Yongxin Chen. Flow matching for averaged systems.arXiv preprint arXiv:2510.14514, 2025

    arXiv 2025

  6. [2]

    Schrödinger bridges of averaged systems.IEEE Transactions on Automatic Control, 2026

    Daniel Owusu Adu and Yongxin Chen. Schrödinger bridges of averaged systems.IEEE Transactions on Automatic Control, 2026

    2026

  7. [4]

    Andrei Agrachev, Davide Barilari, and Ugo Boscain.A Comprehensive Introduction to Sub-Riemannian Geometry, volume

  8. [5]

    A computational fluid mechanics solution to the Monge–Kantorovich mass transfer problem.Numerische Mathematik, 84(3):375–393, 2000

    Jean-David Benamou and Yann Brenier. A computational fluid mechanics solution to the Monge–Kantorovich mass transfer problem.Numerische Mathematik, 84(3):375–393, 2000

    2000

Show all 29 references
  1. [6]

    Georgiy A Bondar, Asmaa Eldesoukey, Yongxin Chen, and Abhishek Halder. Nonlinear non-gaussian density steering with input and noise channel mismatch: Sinkhorn with memory for solving the control-affine Schrödinger bridge problem.arXiv preprint arXiv:2604.23370, 2026

    2026 arXiv

  2. [7]

    Reflected Schrödinger bridge: Density control with path constraints

    Kenneth F Caluya and Abhishek Halder. Reflected Schrödinger bridge: Density control with path constraints. In2021 American Control Conference (ACC), pages 1137–1142. IEEE, 2021

    2021

  3. [8]

    Wasserstein proximal algorithms for the Schrödinger bridge problem: Density control with nonlinear drift.IEEE Transactions on Automatic Control, 67(3):1163–1178, 2021

    Kenneth F Caluya and Abhishek Halder. Wasserstein proximal algorithms for the Schrödinger bridge problem: Density control with nonlinear drift.IEEE Transactions on Automatic Control, 67(3):1163–1178, 2021

    2021

  4. [11]

    Obliquely reflected backward stochastic differential equations

    Jean-François Chassagneux and Adrien Richou. Obliquely reflected backward stochastic differential equations. 2020

    2020

  5. [12]

    Flow matching on general geometries

    Ricky TQ Chen and Yaron Lipman. Flow matching on general geometries. InInternational Conference on Learning Representations, volume 2024, pages 47922–47945, 2024

    2024

  6. [13]

    On the relation between optimal transport and Schrödinger bridges: A stochastic control viewpoint.Journal of Optimization Theory and Applications, 169(2):671–691, 2016

    Yongxin Chen, Tryphon T Georgiou, and Michele Pavon. On the relation between optimal transport and Schrödinger bridges: A stochastic control viewpoint.Journal of Optimization Theory and Applications, 169(2):671–691, 2016

    2016

  7. [14]

    A stochastic control approach to reciprocal diffusion processes.Applied Mathematics and Optimization, 23(1):313–329, 1991

    Paolo Dai Pra. A stochastic control approach to reciprocal diffusion processes.Applied Mathematics and Optimization, 23(1):313–329, 1991

    1991

  8. [15]

    Reflected Schrödinger bridge for constrained generative modeling.arXiv preprint arXiv:2401.03228, 2024

    Wei Deng, Yu Chen, Nicole Tianjiao Yang, Hengrong Du, Qi Feng, and Ricky TQ Chen. Reflected Schrödinger bridge for constrained generative modeling.arXiv preprint arXiv:2401.03228, 2024

    2024 arXiv

  9. [16]

    El Karoui, C

    N. El Karoui, C. Kapoudjian, E. Pardoux, S. Peng, and M. C. Quenez. Reflected solutions of backward sdes, and related obstacle problems for pdes.Annals of Probability, 25(2):702–737, 1997. 10 DANIEL OWUSU ADU

    1997

  10. [17]

    Résolution d’un Système d’équations de M

    Robert Fortet. Résolution d’un Système d’équations de M. Schrödinger.Journal de Mathématiques Pures et Appliquées, 19(1–4):83–105, 1940

    1940

  11. [18]

    Hypoelliptic second-order differential equations.Acta Mathematica, 119:147–171, 1967

    Lars Hörmander. Hypoelliptic second-order differential equations.Acta Mathematica, 119:147–171, 1967

    1967

  12. [19]

    On excursions of reflecting brownian motion.Transactions of the American Mathematical Society, 296(1):239–264, 1986

    Pei Hsu. On excursions of reflecting brownian motion.Transactions of the American Mathematical Society, 296(1):239–264, 1986

    1986

  13. [20]

    Elsevier, 2014

    Nobuyuki Ikeda and Shinzo Watanabe.Stochastic differential equations and diffusion processes, volume 24. Elsevier, 2014

    2014

  14. [21]

    A pde-constrained optimization approach to optimal trajectory planning under uncertainty via reflected Schrödinger bridges.arXiv preprint arXiv:2511.14355, 2025

    Dante Kalise and Wenxin Liu. A pde-constrained optimization approach to optimal trajectory planning under uncertainty via reflected Schrödinger bridges.arXiv preprint arXiv:2511.14355, 2025

    2025 arXiv

  15. [23]

    Reflected diffusion models

    Aaron Lou and Stefano Ermon. Reflected diffusion models. InInternational Conference on Machine Learning, pages 22675–22701. PMLR, 2023

    2023

  16. [24]

    Stochastic Differential Equations

    Bernt Øksendal. Stochastic Differential Equations. InStochastic Differential Equations: An Introduction with Applications, pages 38–50. Springer, 2003

    2003

  17. [25]

    Universitätsverlag Potsdam, 2014

    Andrey Pilipenko.An Introduction to Stochastic Differential Equations with Reflection. Universitätsverlag Potsdam, 2014

    2014

  18. [26]

    Obliquely reflected backward stochastic differential equations.Annales de l’Institut Henri Poincaré, Probabilités et Statistiques, 56(4):2501–2528, 2020

    Adrien Richou. Obliquely reflected backward stochastic differential equations.Annales de l’Institut Henri Poincaré, Probabilités et Statistiques, 56(4):2501–2528, 2020

    2020

  19. [27]

    Verlag der Akademie der Wissenschaften in Kommission bei Walter De Gruyter u ..., 1931

    Erwin Schrödinger.Über die Umkehrung der Naturgesetze. Verlag der Akademie der Wissenschaften in Kommission bei Walter De Gruyter u ..., 1931

    1931

  20. [28]

    On boundary conditions for multidimensional diffusion processes.Theory of Probability & Its Applications, 4(2):164–177, 1959

    Alexander D Venttsel’. On boundary conditions for multidimensional diffusion processes.Theory of Probability & Its Applications, 4(2):164–177, 1959

    1959

  21. [181]

    Cambridge University Press, 2019

    2019

Pith tools

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