REVIEW 4 major objections 4 minor 69 references
Proximal optimal transport divergences
T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash
Pith's one-line read The proximal optimal transport divergence interpolates between optimal transport distances and information divergences through an infimal convolution, and the paper proves its dual variational form and a transport proximal operator.
desk verdict Proximal OT divergences is a useful unification with a correct duality theorem, but the proof of general dual-maximizer existence has a truncation error that is repairable. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the transport proximal operator $\operatorname{prox}^c_{\varepsilon D(\cdot\parallel Q)}(P)=\arg\min_R\{T_c(P,R)+\varepsilon D(R\parallel Q)\}$, whose unique output $R^*$ is the intermediate measure. The proof machinery is convex duality: the paper swaps the infimum over $R$ and the supremum in the Kantorovich dual of $T_c$, then applies the Gibbs variational principle $-\varepsilon\log\mathbb{E}_Q[e^{-\psi/\varepsilon}]=\inf_R\{\mathbb{E}_R[\psi]+\varepsilon D_{\mathrm{KL}}(R\parallel Q)\}$, producing the dual representation (3.8). Everything downstream, including the neural network algorithm, the first variation, and the particle gradient flow, runs on that representation.
What would settle it
Take $P=\delta_0$, $Q=\mathcal{N}(0,1)$, $\varepsilon=1$, and $c(x,y)=|x-y|^2$, and let $\rho=\delta_1-\delta_0$; compute the one-sided derivative $\lim_{\alpha\downarrow 0}[D^c_{\mathrm{KL},1}(P+\alpha\rho\parallel Q)-D^c_{\mathrm{KL},1}(P\parallel Q)]/\alpha$ and compare it with $\int\hat{\phi}^*\,d\rho$ using the dual optimizer $\hat{\phi}^*$. A disagreement would refute Theorem 4.1; a simpler boundary check is that with total variation in place of KL and $P=\delta_0,Q=\delta_1$, the proximal minimizer is non-unique, so Theorem 2.4 cannot hold without the strict-convexity assumption.
Extended reading notes
Core claim
The central claim is that $D^c_\varepsilon(P\parallel Q)=\inf_{R}\{T_c(P,R)+\varepsilon D(R\parallel Q)\}$ is itself a divergence, and that for KL divergence it satisfies the duality $D^c_{\mathrm{KL},\varepsilon}(P\parallel Q)=\sup_{(\phi,\psi)\in\Phi_c}\{\mathbb{E}_P[\phi]-\varepsilon\log\mathbb{E}_Q[e^{-\psi/\varepsilon}]\}$ with $\Phi_c=\{(\phi,\psi):\phi(x)+\psi(y)\le c(x,y)\}$. From this dual identity the paper derives the existence and uniqueness of the intermediate minimizer $R^*$, the Gibbs form $dR^*/dQ=e^{-\psi^*/\varepsilon}/\mathbb{E}_Q[e^{-\psi^*/\varepsilon}]$, and the first-variation formula $\delta D^c_{\mathrm{KL},\varepsilon}/\delta P=\hat{\phi}^*$, the $c$-transform of the optimal dual potential. These results are what turn the divergence into a computable objective and a gradient flow.
Load-bearing premise
The load-bearing premise is that the information divergence $D$ has weakly precompact, strictly convex sublevel sets (Assumption A-D), which secures existence and uniqueness of the intermediate measure $R^*$; for common divergences beyond KL this can fail.
Editorial extensions
If this is right
- Because $D^c_\varepsilon(P\parallel Q)\le \min\{T_c(P,Q),\varepsilon D(P\parallel Q)\}$, the divergence is finite in cases where KL diverges, so it can compare singular or manifold-supported distributions without a density ratio.
- The dual form makes the divergence a concave maximization over transport potentials, so it can be evaluated by convex programming for discrete measures and by input-convex neural networks in continuous spaces.
- As $\varepsilon\to 0$ the rescaled divergence $(1/\varepsilon)D^c_\varepsilon$ increases to $D(P\parallel Q)$, while as $\varepsilon\to\infty$ it increases to $T_c(P,Q)$; thus $\varepsilon$ is a genuine dial between reweighting and transport.
- The dynamic formulation is a first-order mean-field game with a backward Hamilton-Jacobi equation coupled to a forward continuity equation, giving a flow-based picture of the proximal minimizer as the terminal distribution of an optimal control problem.
- The first variation exists along singular perturbations and equals the $c$-transform of the optimal potential, which yields gradient descent algorithms in probability space and particle generative schemes.
Reading between the lines
- Because the dual proof uses only the Gibbs variational principle and Kantorovich duality, the same construction should extend to any divergence with a Donsker-Varadhan-style representation, such as R\'enyi or $\alpha$-divergences; the paper states the extension to $f$-divergences but proves the KL case in full.
- The additivity of the divergence for product measures suggests a tensorized computational strategy in high dimensions: evaluate one-dimensional proximal divergences per coordinate block and sum them, which would make the objective practical for factorized variational inference.
- Interpreting $R^*$ as the output of a transport map followed by an exact Gibbs reweighting suggests a testable regularization for generative flows: enforce that the learned pushforward ends at a Gibbs density relative to the target, rather than only matching samples.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces a family of discrepancies between probability measures, called proximal optimal transport divergences, defined by infimal convolution of an optimal transport cost T_c(P,R) with an information divergence D(R||Q), controlled by a parameter ε. The main theoretical contributions are: elementary properties (finiteness, convexity, lower semicontinuity, uniqueness of the proximal minimizer), interpolation limits as ε→0 and ε→∞, a dual variational representation for the KL case, existence and characterization of dual optimizers, a data-processing inequality, additivity for product measures, a first-variation formula for the divergence, a dynamic Benamou–Brenier / mean-field-game formulation, and closed-form Gaussian examples. The paper also connects the construction to generative modeling algorithms, gradient flows, and input-convex neural network parameterizations.
Significance. If the main results stand, the paper provides a useful unifying framework that interpolates between information divergences and optimal transport, with potential applications in generative modeling, distributionally robust optimization, and gradient flows. The explicit dual representation (Theorem 3.1) and the transport-and-reweighting interpretation of the proximal minimizer are attractive and likely to be used by practitioners. The paper is also commendable for including detailed proofs of the central duality theorem, explicit Gaussian computations, and concrete algorithmic reformulations. However, the proof of dual-maximizer existence contains a sign/truncation error that currently undermines Theorem 3.3(c) and Theorem 4.1; this is repairable but requires a substantive revision of Appendix C.
major comments (4)
- [Appendix C, Theorem C.6, Eqs. (C.39)–(C.43)] The truncation argument proving existence of a dual maximizer is invalid as written. With f_n = \tilde a - φ_n ≥ 0 and g_n = \tilde b - ψ_n ≥ 0, the manuscript defines f_n^{(l)} = max{f_n,l} and g_n^{(l)} = max{g_n,l}. Since max{f_n,l} ≥ f_n, the truncated potentials φ_n^{(l)} = \tilde a - f_n^{(l)} and ψ_n^{(l)} = \tilde b - g_n^{(l)} are pointwise ≤ φ_n and ≤ ψ_n. The dual objective is monotone increasing in (φ,ψ), so inequality (C.43) has the wrong sign: the truncated objective is a lower bound, not an upper bound, for the original objective. Consequently the weak-convergence and monotone-convergence steps leading to (C.47) do not establish the existence of (φ*,ψ*). The intended construction is almost certainly f_n^{(l)} = min{f_n,l} (and similarly for g_n), and with that replacement the argument likely goes through; but as written, Theorem 3.3(c) and the first-variation formula in Theorem 4.1, which both rely on this maximizer, are not proven.
- [Appendix E, proof of Theorem 4.1] The proof of Theorem 4.1 uses the boundedness of the cost c at a load-bearing point, saying “Since c is bounded, arguing as in Theorem C.4,” but Theorem 4.1 assumes only that c is uniformly continuous. Uniformly continuous costs on noncompact Polish spaces need not be bounded (for example c(x,y)=|x−y| on R), and in that case the asserted equiboundedness of the extended potentials \hat φ_α and ψ_α is not justified. The theorem should either add a boundedness hypothesis or provide a localization argument that controls the potentials without global boundedness.
- [Section 3.3, Theorem 3.6] The data-processing inequality is stated as a theorem but its proof is omitted: Appendix D says only that it “follows by a similar argument and is therefore omitted.” This is not a proof of the stated result. The inequality involves the averaged cost Kc and the kernel-induced measures K#P and K#Q, and it is not immediately evident that the dual representation interacts with the kernel in the claimed way. A complete proof, or an explicit statement that this is a conjecture/observation, is needed before this result can be used as a theorem of the paper.
- [Section 2.2, Assumption A-D and Theorem 2.4] The general results in Section 2 are formulated for arbitrary information divergences under Assumption A-D, which requires the level sets {R : D(R||Q) ≤ α} to be precompact and strictly convex. The paper states that these assumptions are satisfied by KL, but for common f-divergences (for example α-divergences with α>1, or total variation) strict convexity or weak precompactness of level sets can fail without additional moment or compactness control. Since Theorem 2.4, the uniqueness of the proximal minimizer R*, and Definition 2.5 are stated in this claimed generality, the scope of the theory is broader than is justified. The main KL-based claims are unaffected, but the framework should either restrict the generality or verify Assumption A-D for each divergence family claimed.
minor comments (4)
- [Appendix C, Eq. (C.42)] In the display defining ψ_n^{(l)}, the right-hand side reads \tilde b - f_n^{(l)}; it should be \tilde b - g_n^{(l)}.
- [Appendix F, Theorem F.2, Eq. (F.11)] The formula for m_R contains the fragment “µ_2+” which appears to be a typographical artifact; please clarify the intended expression.
- [Section 5.2, Eq. (5.5)–(5.6)] The notation \tilde φ and \hat φ is introduced in the text but the displayed equations are not fully consistent about which objects are convex and which are Legendre transforms; please standardize the notation.
- [Section 3.1, Eq. (3.1) and Theorem 3.3(c)] The dual class Φ_c is defined with bounded continuous functions, while Theorem 3.3(c) states that the maximizer lies in L1(P)×L1(Q). The relationship between the two classes should be stated explicitly, since bounded and L1 potentials are not the same set for unbounded costs.
Circularity Check
No load-bearing circularity: the proximal OT dual, optimizer relations, and first variation are derived from standard external duality results; one minor self-citation for an interpolation proof is not load-bearing.
full rationale
The proximal OT divergence is defined independently in Definition 2.1 as an infimal convolution of an OT cost and an information divergence. The central duality Theorem 3.1 is not assumed: its proof (Appendix C) combines the external Kantorovich duality (3.2) and the Gibbs/Donsker-Varadhan variational formula (3.5)-(3.6) with a convex-analysis argument; the resulting representation (3.8) is a theorem, not a restatement of (2.3). The optimizer relations in Theorem 3.3(d) follow from comparing the primal and dual values, using the nonnegativity of the two gap terms in (3.12); they are not fitted or inserted as definitions. Theorem 2.4's uniqueness uses Assumption A-D (precompact, strictly convex divergence level sets), which for KL is an external fact about relative entropy, not the divergence under study. The interpolation Theorem 2.8 is proved in Appendix B; the main text notes the proof is 'essentially the same' as in [7,20], a self-citation that includes two of the present authors ([7]), but because a full proof is supplied and the theorem is not the load-bearing duality step, this is a minor non-load-bearing self-citation rather than circularity. The dynamic formulation (Theorem 2.9) is a direct substitution of the Benamou-Brenier formula (2.18) into (2.3), not a redefinition. No fitted parameters are relabeled as predictions, and no known empirical pattern is simply renamed. A separate correctness concern, distinct from circularity: in Appendix C, Theorem C.6, the truncation f_n^(l)=max{f_n,l} makes the truncated potentials pointwise no larger than (φ_n,ψ_n), so the monotonicity direction claimed in (C.43) appears reversed; as written the existence proof for the dual maximizer needs repair. This is a proof gap, not a circularity, and should be weighed as a correctness risk rather than raising the circularity score beyond 2.
Assumptions & free parameters
assumptions (4)
- domain assumption Assumption A-T: the transport cost T_c(P,Q) is a divergence, and (P,Q) -> T_c(P,Q) is convex and lower semicontinuous in the weak topology.
- domain assumption Assumption A-D: the information divergence D is jointly convex and lower semicontinuous, and level sets {R : D(R||Q) <= alpha} are weakly precompact and strictly convex.
- domain assumption For the duality theorem, the cost c must be bounded below, lower semicontinuous, and satisfy c(x,y) <= a(x)+b(y) with a in L1(P), b in L1(Q).
- standard math Benamou-Brenier formulation holds: for p > 1 and measures with finite p-th moments, W_p^p(P,R) equals the infimum over (v,rho) satisfying the continuity equation.
Cite this review
Pith. "Pith review of Proximal optimal transport divergences." pith.science (2026). https://pith.science/paper/YY73AT3F
@misc{pith2026250512097,
author = {Pith},
title = {Pith review of: Proximal optimal transport divergences},
year = {2026},
howpublished = {\url{https://pith.science/paper/YY73AT3F}},
note = {Machine review of arXiv:2505.12097}
}
read the original abstract
We introduce the proximal optimal transport divergence, a novel discrepancy measure that interpolates between information divergences and optimal transport distances via an infimal convolution formulation. This divergence provides a principled foundation for optimal transport proximals and proximal optimization methods frequently used in generative modeling. We explore its mathematical properties, including smoothness, boundedness, and computational tractability, and establish connections to primal-dual formulations and adversarial learning. The proximal operator associated with the proximal optimal transport divergence can be interpreted as a transport map that pushes a reference distribution toward the optimal generative distribution, which approximates the target distribution that is only accessible through data samples. Building on the Benamou-Brenier dynamic formulation of classical optimal transport, we also establish a dynamic formulation for proximal OT divergences. The resulting dynamic formulation is a first order mean-field game whose optimality conditions are governed by a pair of nonlinear partial differential equations: a backward Hamilton-Jacobi equation and a forward continuity equation. Our framework generalizes existing approaches while offering new insights and computational tools for generative modeling, distributionally robust optimization, and gradient-based learning in probability spaces.
Figures
Reference graph
Works this paper leans on
-
[53]
D´ avid Terj´ ek. Moreau-Yosidaf-divergences. In International Conference on Machine Learn- ing, pages 10214–10224. PMLR, 2021
work page 2021
-
[1]
Building normalizing flows with stochastic interpolants
Michael Samuel Albergo and Eric Vanden-Eijnden. Building normalizing flows with stochastic interpolants. In The Eleventh International Conference on Learning Representations , 2023
2023
-
[2]
Springer Science & Business Media, 2005
Luigi Ambrosio, Nicola Gigli, and Giuseppe Savar´ e.Gradient flows: in metric spaces and in the space of probability measures. Springer Science & Business Media, 2005
2005
-
[3]
Wasserstein generative adversarial networks
Martin Arjovsky, Soumith Chintala, and L´ eon Bottou. Wasserstein generative adversarial networks. In Doina Precup and Yee Whye Teh, editors, Proceedings of the 34th International Conference on Machine Learning , volume 70 of Proceedings of Machine Learning Research, pages 214–223. PMLR, 06–11 Aug 2017
work page 2017
-
[4]
Condi- tional sampling with monotone gans: From generative models to likelihood-free inference
Ricardo Baptista, Bamdad Hosseini, Nikola B Kovachki, and Youssef M Marzouk. Condi- tional sampling with monotone gans: From generative models to likelihood-free inference. SIAM/ASA Journal on Uncertainty Quantification , 12(3):868–900, 2024
work page 2024
-
[5]
A computational fluid mechanics solution to the Monge-Kantorovich mass transfer problem
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
-
[6]
Yoshua Bengio, Nicolas Roux, Pascal Vincent, Olivier Delalleau, and Patrice Marcotte. Con- vex neural networks. Advances in neural information processing systems , 18, 2005
work page 2005
-
[7]
(f-Γ)-divergences: Interpolating between f-divergences and integral probability metrics
Jeremiah Birrell, Paul Dupuis, Markos A Katsoulakis, Yannis Pantazis, and Luc Rey-Bellet. (f-Γ)-divergences: Interpolating between f-divergences and integral probability metrics. Jour- nal of Machine Learning Research , 23(39):1–70, 2022
work page 2022
Show all 69 references
-
[8]
Adversarially robust deep learning with optimal-transport-regularized divergences, 2023
Jeremiah Birrell and Mohammadreza Ebrahimi. Adversarially robust deep learning with optimal-transport-regularized divergences, 2023
2023
-
[9]
Katsoulakis, Luc Rey-Bellet, and Wei Zhu
Jeremiah Birrell, Markos A. Katsoulakis, Luc Rey-Bellet, and Wei Zhu. Structure-preserving GANs. In Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesv´ ari, Gang Niu, and Sivan Sabato, editors, International Conference on Machine Learning, ICML 2022, 17- 23 July 2022...
2022
-
[10]
Function-space regularized R´ enyi divergences
Jeremiah Birrell, Yannis Pantazis, Paul Dupuis, Luc Rey-Bellet, and Markos Katsoulakis. Function-space regularized R´ enyi divergences. InThe Eleventh International Conference on Learning Representations, 2023
2023
-
[11]
Unifying distributionally robust optimization via optimal transport theory
Jose Blanchet, Daniel Kuhn, Jiajin Li, and Bahar Taskesen. Unifying distributionally robust optimization via optimal transport theory. arXiv preprint arXiv:2308.05414 , 2023
2023
-
[12]
´Editions Dunod, Paris, 1999
Ha¨ ım Brezis.Analyse fonctionnelle : th´ eorie et applications. ´Editions Dunod, Paris, 1999
1999
-
[13]
Primal dual methods for wasser- stein gradient flows
Jos´ e A Carrillo, Katy Craig, Li Wang, and Chaozhen Wei. Primal dual methods for wasser- stein gradient flows. Foundations of Computational Mathematics , pages 1–55, 2022. PROXIMAL OPTIMAL TRANSPORT DIVERGENCES 25
2022
-
[14]
Ricky T. Q. Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. Neural ordi- nary differential equations. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa- Bianchi, and R. Garnett, editors, Advances in Neural Information Processing Systems , vol- ume 31....
2018
-
[15]
On the relation between optimal transport and Schr¨ odinger bridges: A stochastic control viewpoint.Journal of Optimization Theory and Applications, 169:671–691, 2016
Yongxin Chen, Tryphon T Georgiou, and Michele Pavon. On the relation between optimal transport and Schr¨ odinger bridges: A stochastic control viewpoint.Journal of Optimization Theory and Applications, 169:671–691, 2016
2016
-
[16]
Robust gener- ative learning with Lipschitz-regularizedα-divergences allows minimal assumptions on target distributions
Ziyu Chen, Hyemin Gu, Markos A Katsoulakis, Luc Rey-Bellet, and Wei Zhu. Robust gener- ative learning with Lipschitz-regularizedα-divergences allows minimal assumptions on target distributions. arXiv preprint arXiv:2405.13962 , 2024
2024 arXiv
-
[17]
McDonald, Wenbo Wu, Shubham Tripathi, Cristina Guarducci, Douglas Russo, Daniel L
Yu-Chen Cheng, Hyemin Gu, Thomas O. McDonald, Wenbo Wu, Shubham Tripathi, Cristina Guarducci, Douglas Russo, Daniel L. Abravanel, Madeline Bailey, Yue Wang, Yun Zhang, Yannis Pantazis, Herbert Levine, Rinath Jeselsohn, Markos A. Katsoulakis, and Franziska Michor. Profet predic...
2025
-
[18]
Wasserstein gradient flows: Appli- cations
Sinho Chewi, Jonathan Niles-Weed, and Philippe Rigollet. Wasserstein gradient flows: Appli- cations. In Statistical Optimal Transport: ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour XLIX– 2019, pages 151–186. Springer, 2025
2019
-
[19]
Dupuis and R.S
P. Dupuis and R.S. Ellis. A Weak Convergence Approach to the Theory of Large Deviations . John Wiley & Sons, New York, 1997
1997
-
[20]
Formulation and properties of a divergence used to com- pare probability measures without absolute continuity
Paul Dupuis and Yixiang Mao. Formulation and properties of a divergence used to com- pare probability measures without absolute continuity. ESAIM: Control, Optimisation and Calculus of Variations, 28:10, 2022
2022
-
[21]
Furio Ercolessi and J. B. Adams. Interatomic potentials from first-principles calculations: the force-matching method. Europhysics Letters, 26(8):583–588, 1994
1994
-
[22]
How to train your neural ODE: the world of Jacobian and kinetic regularization, 2020
Chris Finlay, J¨ orn-Henrik Jacobsen, Levon Nurbekyan, and Adam M Oberman. How to train your neural ODE: the world of Jacobian and kinetic regularization, 2020
2020
-
[23]
Kale flow: A relaxed KL gradient flow for probabilities with disjoint support
Pierre Glaser, Michael Arbel, and Arthur Gretton. Kale flow: A relaxed KL gradient flow for probabilities with disjoint support. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P.S. Liang, and J. Wortman Vaughan, editors, Advances in Neural Information Processing Systems , volume 3...
2021
-
[24]
Generative adversarial nets
Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. Advances in neural information processing systems, 27, 2014
2014
-
[25]
Lipschitz-regularized gradient flows and generative particle algorithms for high-dimensional scarce data
Hyemin Gu, Panagiota Birmpa, Yannis Pantazis, Luc Rey-Bellet, and Markos A Katsoulakis. Lipschitz-regularized gradient flows and generative particle algorithms for high-dimensional scarce data. SIAM Journal on Mathematics of Data Science , 6(4):1205–1235, 2024
2024
-
[26]
Combining Wasserstein-1 and Wasserstein-2 proximals: robust manifold learning via well-posed genera- tive flows
Hyemin Gu, Markos A Katsoulakis, Luc Rey-Bellet, and Benjamin J Zhang. Combining Wasserstein-1 and Wasserstein-2 proximals: robust manifold learning via well-posed genera- tive flows. arXiv preprint arXiv:2407.11901 , 2024
2024 arXiv
-
[27]
A primer on PAC-Bayesian learning
Benjamin Guedj. A primer on PAC-Bayesian learning. arXiv preprint arXiv:1901.05353 , 2019
1901 arXiv
-
[28]
The variational formulation of the fokker–planck equation
Richard Jordan, David Kinderlehrer, and Felix Otto. The variational formulation of the fokker–planck equation. SIAM journal on mathematical analysis , 29(1):1–17, 1998
1998
-
[29]
Mean field games
Jean-Michel Lasry and Pierre-Louis Lions. Mean field games. Japanese journal of mathemat- ics, 2(1):229–260, 2007
2007
-
[30]
W. Li, S. Liu, and S. Osher. A kernel formula for regularized Wasserstein proximal operators. Research in Mathematical Sciences, 10:43, 2023
2023
-
[31]
Flow matching for generative modeling
Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matthew Le. Flow matching for generative modeling. In The Eleventh International Conference on Learning Representations, 2023
2023
-
[32]
Caterini, and Jesse C
Gabriel Loaiza-Ganem, Brendan Leigh Ross, Rasa Hosseinzadeh, Anthony L. Caterini, and Jesse C. Cresswell. Deep generative models through the lens of the manifold hypothesis: A survey and new connections, 2024
2024
-
[33]
Optimal transport map- ping via input convex neural networks
Ashok Makkuva, Amirhossein Taghvaei, Sewoong Oh, and Jason Lee. Optimal transport map- ping via input convex neural networks. In International Conference on Machine Learning , pages 6672–6681. PMLR, 2020. 26 R. BAPTISTA, P. BIRMPA, M. A. KATSOULAKIS, L. REY-BELLET, AND B. J. ZHANG
2020
-
[34]
Interacting particle solutions of fokker–planck equations through gradient–log–density estimation
Dimitra Maoutsa, Sebastian Reich, and Manfred Opper. Interacting particle solutions of fokker–planck equations through gradient–log–density estimation. Entropy, 22(8), 2020
2020
-
[35]
Spectral normal- ization for generative adversarial networks
Takeru Miyato, Toshiki Kataoka, Masanori Koyama, and Yuichi Yoshida. Spectral normal- ization for generative adversarial networks. 02 2018
2018
-
[36]
Wainwright, and Michael I
XuanLong Nguyen, Martin J. Wainwright, and Michael I. Jordan. Estimating divergence functionals and the likelihood ratio by convex risk minimization. IEEE Trans. Inf. Theor. , 56(11):5847–5861, nov 2010
2010
-
[37]
Noid, Ryan J
William G. Noid, Ryan J. Szukalo, Katherine M. Kidder, and Maria C. Lesniewski. Rigorous progress in coarse-graining. Annual Review of Physical Chemistry , 75(1):21–45, June 2024
2024
-
[38]
F-GAN: Training Generative Neural Samplers Using Variational Divergence Minimization
Sebastian Nowozin, Botond Cseke, and Ryota Tomioka. F-GAN: Training Generative Neural Samplers Using Variational Divergence Minimization. InProceedings of the 30th International Conference on Neural Information Processing Systems , NIPS’16, page 271–279, Red Hook, NY, USA, 201...
2016
-
[39]
On the martingale Schr¨ odinger bridge between two dis- tributions
Marcel Nutz and Johannes Wiesel. On the martingale Schr¨ odinger bridge between two dis- tributions. arXiv preprint arXiv:2401.05209 , 2024
2024 arXiv
-
[40]
OT-flow: Fast and accurate continuous normalizing flows via optimal transport
Derek Onken, Samy Wu Fung, Xingjian Li, and Lars Ruthotto. OT-flow: Fast and accurate continuous normalizing flows via optimal transport. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 35, pages 9223–9232, 2021
2021
-
[41]
Proximal algorithms
Neal Parikh and Stephen Boyd. Proximal algorithms. Foundations and Trends in Optimiza- tion, 1(3):127–239, 2014
2014
-
[42]
Computational optimal transport: With applications to data science
Gabriel Peyr´ e, Marco Cuturi, et al. Computational optimal transport: With applications to data science. Foundations and Trends® in Machine Learning, 11(5-6):355–607, 2019
2019
-
[43]
Score-based generative models detect manifolds
Jakiw Pidstrigach. Score-based generative models detect manifolds. Advances in Neural In- formation Processing Systems, 35:35852–35865, 2022
2022
-
[44]
Rachev and Ludger R¨ uschendorf
Svetlozar T. Rachev and Ludger R¨ uschendorf. Mass Transportation Problems: Volume I: Theory; Volume II: Applications . Probability and Its Applications. Springer, 1998. Two- volume set
1998
-
[45]
Frameworks and results in distributionally robust optimization
Hamed Rahimian and Sanjay Mehrotra. Frameworks and results in distributionally robust optimization. Open Journal of Mathematical Optimization , 3:1–85, 2022
2022
-
[46]
Variational inference with normalizing flows
Danilo Rezende and Shakir Mohamed. Variational inference with normalizing flows. In In- ternational conference on machine learning , pages 1530–1538. PMLR, 2015
2015
-
[47]
Optimal transport for applied mathematicians , volume 87 of Progress in Nonlinear Differential Equations and their Applications
Filippo Santambrogio. Optimal transport for applied mathematicians , volume 87 of Progress in Nonlinear Differential Equations and their Applications . Birkh¨ auser/Springer, Cham,
-
[48]
{Euclidean, metric, and Wasserstein } gradient flows: an overview
Filippo Santambrogio. {Euclidean, metric, and Wasserstein } gradient flows: an overview. Bulletin of Mathematical Sciences , 7:87–154, 2017
2017
-
[49]
Sliced score matching: A scalable approach to density and score estimation
Yang Song, Sahaj Garg, Jiaxin Shi, and Stefano Ermon. Sliced score matching: A scalable approach to density and score estimation. In Uncertainty in Artificial Intelligence , pages 574–584. PMLR, 2020
2020
-
[50]
Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole
Yang Song, Jascha Narain Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions. ArXiv, abs/2011.13456, 2021
2011 arXiv
-
[51]
A family of nonparametric density estimation algorithms
Esteban G Tabak and Cristina V Turner. A family of nonparametric density estimation algorithms. Communications on Pure and Applied Mathematics , 66(2):145–164, 2013
2013
-
[52]
Density estimation by dual ascent of the log- likelihood
Esteban G Tabak and Eric Vanden-Eijnden. Density estimation by dual ascent of the log- likelihood. 2010
2010
-
[54]
R´ enyi divergence and Kullback-Leibler divergence.IEEE Transactions on Information Theory , 60(7):3797–3820, 2014
Tim Van Erven and Peter Harremos. R´ enyi divergence and Kullback-Leibler divergence.IEEE Transactions on Information Theory , 60(7):3797–3820, 2014
2014
-
[55]
Tighter generali- sation bounds via interpolation, 2024
Paul Viallard, Maxime Haddouche, Umut S ¸im¸ sekli, and Benjamin Guedj. Tighter generali- sation bounds via interpolation, 2024
2024
-
[56]
American Mathematical Society, Providence, RI, 2003
C´ edric Villani.Topics in optimal transportation , volume 58 of Graduate Studies in Mathe- matics. American Mathematical Society, Providence, RI, 2003
2003
-
[57]
Optimal transport , volume 338 of Grundlehren der mathematischen Wis- senschaften [Fundamental Principles of Mathematical Sciences]
C´ edric Villani. Optimal transport , volume 338 of Grundlehren der mathematischen Wis- senschaften [Fundamental Principles of Mathematical Sciences] . Springer-Verlag, Berlin,
-
[58]
Efficient neural network approaches for conditional optimal transport with applica- tions in bayesian inference
Zheyu Oliver Wang, Ricardo Baptista, Youssef Marzouk, Lars Ruthotto, and Deepanshu Verma. Efficient neural network approaches for conditional optimal transport with applica- tions in bayesian inference. arXiv preprint arXiv:2310.16975 , 2023
-
[59]
A mean-field games laboratory for generative modeling
Benjamin J Zhang and Markos A Katsoulakis. A mean-field games laboratory for generative modeling. arXiv preprint arXiv:2304.13534 , 2023
2023 arXiv
-
[60]
Wasserstein proximal operators describe score-based generative models and resolve memo- rization
Benjamin J Zhang, Siting Liu, Wuchen Li, Markos A Katsoulakis, and Stanley J Osher. Wasserstein proximal operators describe score-based generative models and resolve memo- rization. arXiv preprint arXiv:2402.06162 , 2024. Appendix A. Proofs of the divergence theorems in Sectio...
2024 arXiv
-
[63]
+εD(R∗ 3∥Q) < 1 2 (Tc(P,R∗
-
[64]
+ϵD(R∗ 1∥Q) +Tc(P,R∗
-
[65]
This contradicts the fact that R∗ 1 and R∗ 2 are minimizers
+ϵD(R∗ 2∥Q)) = Dc ε(P∥Q). This contradicts the fact that R∗ 1 and R∗ 2 are minimizers. □ Proof of Theorem 2.3.The divergence property, convexity and lower semiconitnuity are proved as follows: Divergence property: For the divergence property the nonnegativity in (2.8) follows ...
-
[66]
BAPTISTA, P
Note also that the function F (P,Q,R ) = 28 R. BAPTISTA, P. BIRMPA, M. A. KATSOULAKIS, L. REY-BELLET, AND B. J. ZHANG TC(P,R ) +εD(R∥Q) is jointly convex. We have then (A.5) Dc ε(αP1 + (1−α)P2∥αQ1 + (1−αQ2) ≤F (α(P1,Q 1,R∗
-
[67]
+ (1−α)(P2,Q 2,R∗ 2)) ≤αF (P1,Q 1,R∗
-
[68]
+ (1−α)F (P2,Q 2,R∗ 2) =αDc ε(P1∥Q1) + (1−α)Dc ε(P2∥Q2) which proves convexity. Lower semicontinuity: For the lower semicontinuity property let ( Pn,Qn) ∈ P(X)×P (Y ) a weakly convergent sequence and suppose nk is such that (A.6) lim inf n Dc ε(Pn∥Qn) = lim k Dc ε(Pnk∥Qnk). If...
-
[69]
and Q =N (m2,σ 2 2). Then the OT-divergence made by 2-Wasserstein and the KL-divergence is given by choosing the intermediary distribution to be the Gaussian distribution R =N (mR,σ 2 R) with mR = m1 + ε 2σ2 2 m2 1 + ε 2σ2 2 (F.1) σR = σ1 + q σ2 1 + 2ε(1 + ε 2σ2 2 ) 2(1 + ε 2σ...
-
[2009]
PROXIMAL OPTIMAL TRANSPORT DIVERGENCES 27
Old and new. PROXIMAL OPTIMAL TRANSPORT DIVERGENCES 27
-
[2015]
Calculus of variations, PDEs, and modeling
Reviewed August 15, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.