Friday, July 14, 2023 - 09:30 in ZiF
Jump-diffusion consensus-based optimization
A talk in the SPDEs, optimal control and mean field games series by
Michael Tretyakov from Nottingham
| Abstract: |
A new consensus based optimization (CBO) method, where an interacting particle
system is driven by jump-diffusion stochastic differential equations, is introduced.
Well-posedness of the particle system and of its mean-field limit is studied.
Convergence of the interacting particle system to the mean-field limit and convergence
of a discretized particle system to the continuous-time dynamics in the mean-square
sense are proved. Convergence of the mean-field (McKean-Vlasov) jump-diffusion
SDEs to global minimizers for a large class of objective functions is also considered.
Numerical tests performed on benchmark objective functions demonstrate improved
performance of the proposed CBO method over earlier CBO models. The talk is
based on a joint work with Dante Kalise (Imperial College) and Akash Sharma (University
of Nottingham).
|
Back
