Papers
Topics
Authors
Recent
Assistant
AI Research Assistant
Well-researched responses based on relevant abstracts and paper content.
Custom Instructions Pro
Preferences or requirements that you'd like Emergent Mind to consider when generating responses.
Gemini 2.5 Flash
Gemini 2.5 Flash 74 tok/s
Gemini 2.5 Pro 55 tok/s Pro
GPT-5 Medium 19 tok/s Pro
GPT-5 High 24 tok/s Pro
GPT-4o 109 tok/s Pro
Kimi K2 212 tok/s Pro
GPT OSS 120B 464 tok/s Pro
Claude Sonnet 4 37 tok/s Pro
2000 character limit reached

Simulating conditioned diffusions on manifolds (2403.05409v3)

Published 8 Mar 2024 in math.PR, math.ST, and stat.TH

Abstract: To date, most methods for simulating conditioned diffusions are limited to the Euclidean setting. The conditioned process can be constructed using a change of measure known as Doob's $h$-transform. The specific type of conditioning depends on a function $h$ which is typically unknown in closed form. To resolve this, we extend the notion of guided processes to a manifold $M$, where one replaces $h$ by a function based on the heat kernel on $M$. We consider the case of a Brownian motion with drift, constructed using the frame bundle of $M$, conditioned to hit a point $x_T$ at time $T$. We prove equivalence of the laws of the conditioned process and the guided process with a tractable Radon-Nikodym derivative. Subsequently, we show how one can obtain guided processes on any manifold $N$ that is diffeomorphic to $M$ without assuming knowledge of the heat kernel on $N$. We illustrate our results with numerical simulations of guided processes and Bayesian parameter estimation based on discrete-time observations. For this, we consider both the torus and the Poincar\'e disk.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (40)
  1. Generalization of a retard Gronwall-like inequality and its applications. Applied Mathematics and Computation, 165(06), 599–612.
  2. Diffusion Bridges for Stochastic Hamiltonian Systems and Shape Evolutions. SIAM Journal on Imaging Sciences, 15(1), 293–323.
  3. Manifolds.jl: An Extensible Julia Framework for Data Analysis on Manifolds. arXiv:2106.08777.
  4. Baudoin, Fabrice. 2002. Conditioned stochastic differential equations: theory, examples and application to finance. Stochastic Processes and their Applications, 100(1), 109–145.
  5. Baudoin, Fabrice. 2004. Conditioning and initial enlargement of filtration on a Riemannian manifold. Annals of probability, 2286–2303.
  6. Exact and computationally efficient likelihood-based estimation for discretely observed diffusion processes. J. R. Stat. Soc. Ser. B Stat. Methodol., 68(3), 333–382. With discussions and a reply by the authors.
  7. Retrospective exact simulation of diffusion sample paths with applications. Bernoulli, 12(6), 1077–1098.
  8. MCMC Methods for diffusion bridges. Stochastics and Dynamics, 08(03), 319–350.
  9. Simulation of elliptic and hypo-elliptic conditional diffusions. Advances in Applied Probability, 52(1), 173–212.
  10. A piecewise deterministic Monte Carlo method for diffusion bridges.
  11. Simulation of multivariate diffusion bridges. J. R. Stat. Soc. Ser. B. Stat. Methodol., 78(2), 343–369.
  12. Inference for partially observed Riemannian Ornstein–Uhlenbeck diffusions of covariance matrices. Bernoulli, 29(4), 2961–2986.
  13. Clark, J. M. C. 1990. The simulation of pinned diffusions. Pages 1418–1420 of: Decision and Control, 1990., Proceedings of the 29th IEEE Conference on. IEEE.
  14. Corstanje, Marc. 2024. https://github.com/macorstanje/manifold_bridge.jl. Code examples for simulations.
  15. Conditioning continuous-time Markov processes by guiding. Stochastics, 95(6), 963–996.
  16. Simulation of conditioned diffusions and application to parameter estimation. Stochastic Processes and their Applications, 116(11), 1660–1675.
  17. Numerical techniques for maximum likelihood estimation of continuous-time diffusion processes. J. Bus. Econom. Statist., 20(3), 297–338. With comments and a reply by the authors.
  18. Elworthy, David. 1988. Geometric aspects of diffusions on manifolds. Pages 277–425 of: Hennequin, Paul-Louis (ed), École d’Été de Probabilités de Saint-Flour XV–XVII, 1985–87. Berlin, Heidelberg: Springer Berlin Heidelberg.
  19. Emery, Michel. 1989. Stochastic Calculus in Manifolds. Universitext. Berlin, Heidelberg: Springer Berlin Heidelberg.
  20. Learning and Inference in Computational Systems Biology. MIT Press. Chap. Markov chain Monte Carlo algorithms for SDE parameter estimation, pages 253–276.
  21. Sampling Conditioned Diffusions. Pages 159–186 of: Trends in Stochastic Analysis. London Mathematical Society Lecture Note Series, vol. 353. Cambridge University Press.
  22. Diffusion Means and Heat Kernel on Manifolds. Pages 111–118 of: Nielsen, Frank, & Barbaresco, Frédéric (eds), Geometric Science of Information. Cham: Springer International Publishing.
  23. Hsu, Elton P. 2002. Stochastic analysis on manifolds. Vol. 38. American Mathematical Soc.
  24. Simulation of Conditioned Semimartingales on Riemannian Manifolds. arXiv preprint arXiv:2105.13190.
  25. Simulation of conditioned diffusions on the flat torus. Pages 685–694 of: Geometric Science of Information: 4th International Conference, GSI 2019, Toulouse, France, August 27–29, 2019, Proceedings 4. Springer.
  26. Discrete-Time Observations of Brownian Motion on Lie Groups and Homogeneous Spaces: Sampling and Metric Estimation. Algorithms, 15(8), 290.
  27. Computational Anatomy in Theano. Pages 164–176 of: Graphs in Biomedical Image Analysis, Computational Anatomy and Imaging Genetics. Cham: Springer International Publishing.
  28. On generating Monte Carlo samples of continuous diffusion bridges. J. Amer. Statist. Assoc., 105(490), 820–838.
  29. Lovett, Stephen. 2010. Differential geometry of manifolds. AK Peters/CRC Press.
  30. Manton, Jonathan H. 2013. A primer on stochastic differential geometry for signal processing. IEEE Journal of Selected Topics in Signal Processing, 7(4), 681–699.
  31. Continuous-discrete smoothing of diffusions. Electronic Journal of Statistics, 15(2), 4295 – 4342.
  32. A technique for exponential change of measure for Markov processes. Bernoulli, 8(6), 767–785.
  33. Data Augmentation for Diffusions. J. Comput. Graph. Statist., 22(3), 665–688.
  34. Examples of adaptive MCMC. Journal of computational and graphical statistics, 18(2), 349–367.
  35. Guided proposals for simulating multi-dimensional diffusion bridges. Bernoulli, 23(4A), 2917–2950.
  36. Sommer, Stefan. 2020. Probabilistic approaches to geometric statistics: Stochastic processes, transition distributions, and fiber bundle geometry. Pages 377–416 of: Pennec, Xavier, Sommer, Stefan, & Fletcher, Tom (eds), Riemannian Geometric Statistics in Medical Image Analysis. Academic Press.
  37. Bridge Simulation and Metric Estimation on Landmark Manifolds. Pages 79–91 of: Graphs in Biomedical Image Analysis, Computational Anatomy and Imaging Genetics. Lecture Notes in Computer Science. Springer.
  38. Fast communication conditional path sampling of SDEs and the Langevin MCMC method. Commun. Math. Sci., 2(4), 685–697.
  39. Sturm, K.T. 1993. Schrödinger Semigroups on Manifolds. Journal of Functional Analysis, 118(2), 309–350.
  40. Improved bridge constructs for stochastic differential equations. Statistics and Computing, 27(4), 885–900.
Citations (1)

Summary

  • The paper presents a guided simulation method for conditioned diffusion processes on Riemannian manifolds using heat kernel techniques.
  • It establishes theoretical equivalence with a tractable Radon-Nikodym derivative to correct discrepancies between simulated and true processes.
  • Numerical examples, including a torus simulation, demonstrate robust parameter estimation and practical applications in manifold modeling.

Simulating Conditioned Diffusions on Manifolds

The paper "Simulating Conditioned Diffusions on Manifolds" by Corstanje et al. addresses the challenge of simulating diffusion processes conditioned on certain events, specifically bridging processes to target points on manifolds. Traditionally, conditioned simulations have been confined to Euclidean spaces, with limited extensions to manifold structures. The work presented offers a broad expansion into manifold settings, employing a novel approach that leverages the heat kernel to guide the diffusion processes.

Overview and Key Contributions

This paper extends the concept of guided diffusion processes to Riemannian manifolds. Specifically, the authors focus on simulating a manifold diffusion process conditioned to hit a specific point at a predetermined future time. The primary mathematical challenge lies in the intractability of the transition density function, which is ameliorated by introducing a guided process based on known heat kernels on comparable manifolds.

The methodology is grounded in Doob's hh-transform, traditionally employed in constructing conditioned processes. The authors circumvent the unknown hh function by substituting it with a function related to the known heat kernel on the manifold MM. They demonstrate that this guided process retains essential properties necessary for practical simulation.

Theoretical results establish the equivalency of the laws of the conditioned process and the guided process with a tractable Radon-Nikodym derivative. This equivalency is crucial for the simulation, as it allows the use of likelihood ratios to correct for discrepancies between the true and guided processes.

Numerical Results and Practical Applications

The paper illustrates the theoretical framework with simulations, including an insightful example of parameter estimation on a torus. The simulation results verify the effectiveness of the proposed method by demonstrating the ability to model complex manifold processes and provide a practical implementation pathway for researchers in related fields.

Implications and Future Directions

The implications of this research are twofold: theoretical and practical. Theoretically, it opens up new pathways for statistical inference on manifold structures, potentially benefiting various fields such as shape analysis and geometric statistics. Practically, it bridges the gap between theoretical manifold diffusion processes and their application in real-world data modeling scenarios, enhancing computational efficiency in fields like medical imaging and evolutionary biology.

Looking beyond the scope of this paper, future work could explore the extension of these methodologies to other types of conditionings beyond point hit conditions, diversifying the potential applications even further. Additionally, adapting the approach to non-compact or more complex manifolds might unveil new insights into manifold-based diffusion processes.

Conclusion

Corstanje et al.'s paper significantly advances the field of conditioned diffusions by extending simulation capabilities to manifold settings. By utilizing heat kernel knowledge and guided processes, the authors provide a robust framework that addresses longstanding challenges in the simulation and analysis of conditioned stochastic differential equations on manifolds. This work not only contributes valuable academic insights but also offers practical solutions for manifold-related simulations in various scientific domains.

Lightbulb On Streamline Icon: https://streamlinehq.com

Continue Learning

We haven't generated follow-up questions for this paper yet.

List To Do Tasks Checklist Streamline Icon: https://streamlinehq.com

Collections

Sign up for free to add this paper to one or more collections.

X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets

This paper has been mentioned in 4 posts and received 91 likes.

Don't miss out on important new AI/ML research

See which papers are being discussed right now on X, Reddit, and more:

“Emergent Mind helps me see which AI papers have caught fire online.”

Philip

Philip

Creator, AI Explained on YouTube