Seminar: Fast algorithms for simulation and sampling of particle systems
Seminar| Institute of Mathematical Sciences
Time: Friday, May 8th, 2026,15:00-16:00
Location: IMS, RS518
Speaker: Lei Li, Shanghai Jiaotong University
Abstract: We first introduce a particle system approximation based on a diffusion limit for the Landau kinetic equation. Building on this, we propose a simulation algorithm that relies solely on random grouping and sampling via spherical Brownian motion. The algorithm has a per-step computational complexity of O(N), strictly preserves energy, and thus supports long-time simulations. We then present a discontinuous Hamiltonian method for sampling interacting particle systems in the grand canonical ensemble. Compared with the traditional Metropolis method, this approach involves no rejection steps. When combined with random batch method, it achieves O(N) computational complexity per step, and the resulting samples exhibit significantly lower correlation than those generated by traditional methods.