Paperlane

no. 17 · Wednesday 18 November 2026

A Study of the PDE Model for the Anisotropic Consensus-Based Optimization

Hui Huang, Keyu Li and Jinhuan Wang

math.AP · 37 pp

Abstract

We investigate the nonlinear Fokker–Planck equation associated with anisotropic consensus-based optimization. Under suitable assumptions on the objective function, we establish global existence and uniqueness of density-valued weak solutions for bounded initial probability densities with finite fourth moment. The main analytical challenge arises from the component-wise degeneracy of the diffusion, which vanishes along moving coordinate hyperplanes determined by the weighted consensus point. To overcome this difficulty, we develop spatial translation estimates based on weighted gradient bounds and uniform density estimates. Combined with temporal translation and moment estimates, these yield strong compactness in $L^2((0,T)\times\mathbb{R}^d)$ through the Fréchet–Kolmogorov compactness criterion, enabling passage to the limit in the regularized equations. Uniqueness is established through the associated McKean–Vlasov stochastic differential equation and Wasserstein stability of the weighted consensus point. Furthermore, we prove the propagation of $H^2$-regularity for sufficiently regular initial data.

comments
37 pages
fields
Analysis of PDEs, Optimization and Control
pdf ↗ page arXiv:2611.22208 ↗