Numerical computation and estimation of large deviations

Statistical Physics and Complexity Group meeting

Numerical computation and estimation of large deviations

  • Event time: 11:30am until 12:30pm
  • Event date: 1st July 2020
  • Speaker: Professor Hugo Touchette (Stellenbosch University, Department of Mathematical Sciences)

Event details

I will give in this talk a basic overview of numerical methods developed recently for computing large deviation functions characterizing the fluctuations of observables of Markov processes, which are used in statistical physics to model nonequilibrium processes. I will focus on two methods - direct diagonalization and importance sampling - to explain how the estimation of large deviations relates in general to optimal control theory, and what differences or difficulties arise when treating reversible vs non-reversible processes. I will end by discussing pros and cons of different numerical methods, as well as problems that remain in the field of numerical large deviations.