Samplepath Large Deviations: Theory and Numerical Tools

Summary: Rare but extreme events in complex systems can often efficiently be described by samplepath large deviations: In the limit of some smallness-parameter approaching zero (such as temperature for chemical reactions, inverse number of particles for thermodynamic limits, or inverse timescale separation for multiscale systems), probabilities and most likely pathways of occurence can are readily accessible. For large and strongly coupled stochastic systems, such as climate, atmosphere, or ocean, the corresponding computations pose a huge numerical challenge. These method borrow heavily from field theory, and represent the rare probability as a path integral, necessitating the computation of instantons and fluctuation determinants. In this project, we adress these challenges, including (1) how to compute the large deviation minimizer (instanton) for large systems, (2) how to compute next-order prefactor corrections, and (3) how to deal with heavy-tailed distributions

Relevant publications

  1. T. Schorlepp, T. Grafke, and R. Grauer, "Gel'fand-Yaglom type equations for calculating fluctuations around Instantons in stochastic systems", J. Phys. A: Math. Theor. 54 (2021), 235003 (link)

  2. T. Grafke, T. Schäfer, and E. Vanden-Eijnden, "Sharp Asymptotic Estimates for Expectations, Probabilities, and Mean First Passage Times in Stochastic Systems with Small Noise", ArXiv (2021) (link)

  3. M. Alqahtani, and T. Grafke, "Instantons for rare events in heavy-tailed distributions", J. Phys. A: Math. Theor. 54 (2021), 175001 (link)

  4. G. Ferré and T. Grafke, "Approximate Optimal Controls via Instanton Expansion for Low Temperature Free Energy Computation", SIAM Multiscale Model. Simul. 19(3) (2021), 1310 (link)

  5. T. Grafke, and E. Vanden-Eijnden, "Numerical computation of rare events via large deviation theory", Chaos 29 (2019), 063118 (link)

  6. T. Grafke, "String Method for Generalized Gradient Flows: Computation of Rare Events in Reversible Stochastic Processes", J. Stat. Mech. 2019/4 (2019), 043206 (link)