Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
156 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Sinkhorn Flow: A Continuous-Time Framework for Understanding and Generalizing the Sinkhorn Algorithm (2311.16706v1)

Published 28 Nov 2023 in cs.LG, math.PR, and stat.ML

Abstract: Many problems in machine learning can be formulated as solving entropy-regularized optimal transport on the space of probability measures. The canonical approach involves the Sinkhorn iterates, renowned for their rich mathematical properties. Recently, the Sinkhorn algorithm has been recast within the mirror descent framework, thus benefiting from classical optimization theory insights. Here, we build upon this result by introducing a continuous-time analogue of the Sinkhorn algorithm. This perspective allows us to derive novel variants of Sinkhorn schemes that are robust to noise and bias. Moreover, our continuous-time dynamics not only generalize but also offer a unified perspective on several recently discovered dynamics in machine learning and mathematics, such as the "Wasserstein mirror flow" of (Deb et al. 2023) or the "mean-field Schr\"odinger equation" of (Claisse et al. 2023).

Definition Search Book Streamline Icon: https://streamlinehq.com
References (60)
  1. Gradient flows: in metric spaces and in the space of probability measures. Springer Science & Business Media, 2005.
  2. Mirror Descent with Relative Smoothness in Measure Spaces, with application to Sinkhorn and EM, October 2022.
  3. Mirror descent with relative smoothness in measure spaces, with application to sinkhorn and em. Advances in Neural Information Processing Systems, 35:17263–17275, 2022.
  4. Mirror sinkhorn: Fast online optimization on transport polytopes. In International Conference on Machine Learning, pages 1595–1613. PMLR, 2023.
  5. Bregman monotone optimization algorithms. SIAM Journal on control and optimization, 42(2):596–636, 2003.
  6. Mirror descent and nonlinear projected subgradient methods for convex optimization. Operations Research Letters, 31(3):167–175, 2003.
  7. Michel Benaïm. Dynamics of stochastic approximation algorithms. In Jacques Azéma, Michel Émery, Michel Ledoux, and Marc Yor, editors, Séminaire de Probabilités XXXIII, volume 1709 of Lecture Notes in Mathematics, pages 1–68. Springer Berlin Heidelberg, 1999.
  8. Entropic Optimal Transport Solutions of the Semigeostrophic Equations. 2023.
  9. Schrödinger Bridge Samplers. In arXiv preprint arXiv:1912.13170, 2019.
  10. The schrödinger bridge between gaussian measures has a closed form. In International Conference on Artificial Intelligence and Statistics, pages 5802–5833. PMLR, 2023.
  11. Likelihood Training of Schrödinger Bridge using Forward-Backward SDEs Theory. In International Conference on Learning Representations (ICLR), 2022.
  12. Entropic and displacement interpolation: a computational approach using the hilbert metric. SIAM Journal on Applied Mathematics, 76(6):2375–2396, 2016.
  13. Optimal Transport in Systems and Control. Annual Review of Control, Robotics, and Autonomous Systems, 4, 2021.
  14. Gradient estimates for the schrödinger potentials: convergence to the brenier map and quantitative stability. Communications in Partial Differential Equations, pages 1–49, 2023.
  15. A connection between tempering and entropic mirror descent, 2023.
  16. Mean field optimization problem regularized by fisher information. arXiv preprint arXiv:2302.05938, 2023.
  17. Marco Cuturi. Sinkhorn distances: Lightspeed computation of optimal transport. Advances in neural information processing systems, 26, 2013.
  18. Diffusion Schrödinger Bridge with Applications to Score-Based Generative Modeling. In Advances in Neural Information Processing Systems (NeurIPS), volume 35, 2021.
  19. Wasserstein mirror gradient flow as the limit of the sinkhorn algorithm. arXiv preprint arXiv:2307.16421, 2023.
  20. Fast stochastic bregman gradient methods: Sharp analysis and variance reduction. In International Conference on Machine Learning, pages 2815–2825. PMLR, 2021.
  21. Wendell H. Fleming. Exit probabilities and optimal stochastic control. Applied Mathematics and Optimization, 4(1):329–346, March 1977. ISSN 1432-0606. doi: 10.1007/BF01442148.
  22. H Föllmer. An entropy approach to the time reversal of diffusion processes, stochastic differential systems filtering and control (m. metivier, e. pardoux, ed.), 1985.
  23. Robert Fortet. Résolution d’un systeme d’équations de M. Schrödinger. J. Math. Pure Appl. IX, 1, 1940.
  24. On the convergence rate of sinkhorn’s algorithm. arXiv preprint arXiv:2212.06000, 2022.
  25. Paul R Halmos. Measure theory, volume 18. Springer, 2013.
  26. Fastest rates for stochastic mirror descent methods. Computational Optimization and Applications, 79:717–766, 2021.
  27. U. G. Haussmann and E. Pardoux. Time Reversal of Diffusions. The Annals of Probability, 14(4):1188–1205, October 1986. ISSN 0091-1798, 2168-894X. doi: 10.1214/aop/1176992362.
  28. Denoising Diffusion Probabilistic Models. In Advances in Neural Information Processing Systems (NeurIPS), 2020.
  29. Path Integral Stochastic Optimal Control for Sampling Transition Paths. arXiv preprint arXiv:2207.02149, 2022.
  30. Finding mixed nash equilibria of generative adversarial networks. In International Conference on Machine Learning, pages 2810–2819, 2019.
  31. Schrödinger-Föllmer Sampler: Sampling without Ergodicity. arXiv preprint arXiv:2106.10880, 2021.
  32. The variational formulation of the fokker–planck equation. SIAM journal on mathematical analysis, 29(1):1–17, 1998.
  33. A dynamical system view of langevin-based non-convex sampling. arXiv preprint arXiv:2210.13867, 2022a.
  34. The dynamics of Riemannian Robbins-Monro algorithms. In COLT 2022-35th Annual Conference on Learning Theory, pages 1–31, 2022b.
  35. Accelerated mirror descent in continuous and discrete time. In Advances in neural information processing systems, pages 2845–2853, 2015.
  36. Solomon Kullback. Probability densities with given marginals. The Annals of Mathematical Statistics, 39(4):1236–1243, 1968.
  37. Flavien Léger. A gradient descent perspective on sinkhorn. Applied Mathematics & Optimization, 84(2):1843–1855, 2021.
  38. Christian Léonard. Girsanov theory under a finite entropy condition, January 2011.
  39. Christian Léonard. A survey of the Schrödinger problem and some of its connections with optimal transport. Discrete and Continuous Dynamical Systems - Series A, 34(4):1533–1574, 2014.
  40. Deep Generalized Schrödinger Bridge. In Advances in Neural Information Processing Systems (NeurIPS), 2022.
  41. Online sinkhorn: Optimal transport distances from sample streams. Advances in Neural Information Processing Systems, 33:1657–1667, 2020.
  42. A unified stochastic approximation framework for learning in games. Mathematical Programming, pages 1–51, 2023.
  43. Konstantin Mishchenko. Sinkhorn algorithm as a special case of stochastic mirror descent. arXiv preprint arXiv:1909.06918, 2019.
  44. AS Nemirovsky and DB Yudin. Problem complexity and method efficiency in optimization. 1983.
  45. Entropic optimal transport: Convergence of potentials. Probability Theory and Related Fields, 184(1-2):401–424, October 2022. ISSN 0178-8051, 1432-2064. doi: 10.1007/s00440-021-01096-8.
  46. Felix Otto. The geometry of dissipative evolution equations: the porous medium equation. 2001.
  47. Unbalanced diffusion schr\\\backslash\” odinger bridge. arXiv preprint arXiv:2306.09099, 2023.
  48. Computational optimal transport: With applications to data science. Foundations and Trends® in Machine Learning, 11(5-6):355–607, 2019.
  49. Entropic estimation of optimal transport maps. arXiv preprint arXiv:2109.12004, 2021.
  50. On the Markov processes of Schrödinger, the Feynman-Kac formula and stochastic control. In M. A. Kaashoek, J. H. Van Schuppen, and A. C. M. Ran, editors, Realization and Modelling in System Theory, pages 497–504. Birkhäuser Boston, Boston, MA, 1990. ISBN 978-1-4612-8033-0 978-1-4612-3462-3. doi: 10.1007/978-1-4612-3462-3˙55.
  51. R Tyrrell Rockafellar. Convex analysis, volume 11. Princeton university press, 1997.
  52. Concerning nonnegative matrices and doubly stochastic matrices. Pacific Journal of Mathematics, 21(2):343–348, 1967.
  53. Aligned diffusion schr\\\backslash\” odinger bridges. arXiv preprint arXiv:2302.11419, 2023.
  54. Generative Modeling by Estimating Gradients of the Data Distribution. In Advances in Neural Information Processing Systems (NeurIPS), 2019.
  55. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020.
  56. Variational principles for mirror descent and mirror langevin dynamics. IEEE Control Systems Letters, 2023.
  57. Solving Schrödinger Bridges via Maximum Likelihood. Entropy, 23(9), 2021.
  58. Cédric Villani. Optimal transport: old and new, volume 338. Springer Science & Business Media, 2008.
  59. Deep Generative Learning via Schrödinger Bridge. In International Conference on Machine Learning (ICML), 2021.
  60. Path Integral Sampler: A Stochastic Control Approach For Sampling. In International Conference on Learning Representations (ICLR), 2022.
Citations (1)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now