21
5

Bayesian Nonparametric Inference in McKean-Vlasov models

Abstract

We consider nonparametric statistical inference on a periodic interaction potential WW from noisy discrete space-time measurements of solutions ρ=ρW\rho=\rho_W of the nonlinear McKean-Vlasov equation, describing the probability density of the mean field limit of an interacting particle system. We show how Gaussian process priors assigned to WW give rise to posterior mean estimators that exhibit fast convergence rates for the implied estimated densities ρˉ\bar \rho towards ρW\rho_W. We further show that if the initial condition ϕ\phi is not too smooth and satisfies a standard deconvolvability condition, then one can consistently infer the potential WW itself at convergence rates NθN^{-\theta} for appropriate θ>0\theta>0, where NN is the number of measurements. The exponent θ\theta can be taken to approach 1/21/2 as the regularity of WW increases corresponding to `near-parametric' models.

View on arXiv
Comments on this paper