Current browse context:
math.PR
Change to browse by:
References & Citations
Mathematics > Probability
Title: On the mean field limit of the Random Batch Method for interacting particle systems
(Submitted on 24 May 2020 (v1), last revised 21 Nov 2020 (this version, v3))
Abstract: The Random Batch Method proposed in our previous work [Jin et al., J. Comput. Phys., 400(1), 2020] is not only a numerical method for interacting particle systems and its mean-field limit, but also can be viewed as a model of particle system in which particles interact, at discrete time, with randomly selected mini-batch of particles. In this paper we investigate the mean-field limit of this model as the number of particles $N \to \infty$. Unlike the classical mean field limit for interacting particle systems where the law of large numbers plays the role and the chaos is propagated to later times, the mean field limit now does not rely on the law of large numbers and chaos is imposed at every discrete time. Despite this, we will not only justify this mean-field limit (discrete in time) but will also show that the limit, as the discrete time interval $\tau \to 0$, approaches to the solution of a nonlinear Fokker-Planck equation arising as the mean-field limit of the original interacting particle system in Wasserstein distance.
Submission history
From: Lei Li [view email][v1] Sun, 24 May 2020 13:04:46 GMT (945kb,D)
[v2] Wed, 3 Jun 2020 08:39:26 GMT (945kb,D)
[v3] Sat, 21 Nov 2020 09:38:03 GMT (946kb,D)
Link back to: arXiv, form interface, contact.