Abstract:We address the approximation of functionals depending on a system of particles, described by stochastic differential equations (SDEs), in the mean-field limit when the number of particles approaches infinity. This problem is equivalent to estimating the weak solution of the limiting McKean-Vlasov SDE. To that end, our approach uses systems with finite numbers of particles and a time-stepping scheme. In this case, there are two discretization parameters: the number of time steps and the number of particles. Based on these two parameters, we consider different variants of the Monte Carlo and Multilevel Monte Carlo (MLMC) methods and show that, in the best case, the optimal work complexity of MLMC, to estimate the functional in one typical setting with an error tolerance of $\mathrm{TOL}$, is $\mathcal O\left({\mathrm{TOL}^{-3}}\right)$ when using the partitioning estimator and the Milstein time-stepping scheme. We also consider a method that uses the recent Multi-index Monte Carlo method and show an improved work complexity in the same typical setting of $\mathcal O\left(\mathrm{TOL}^{-2}\log(\mathrm{TOL}^{-1})^2\right)$. Our numerical experiments are carried out on the so-called Kuramoto model, a system of coupled oscillators.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to effectively estimate the function values depending on large - scale particle systems when the number of particles tends to infinity. Specifically, the author focuses on the method of estimating the weak solution of the McKean - Vlasov stochastic differential equation under the mean - field limit. To achieve this goal, the author uses a system with a finite number of particles and a time - stepping scheme, and considers different Monte Carlo methods and their multi - level variants, including the Multi - index Monte Carlo (MIMC) method, to improve computational efficiency.
### Problem Background
In many scientific and engineering applications, it is necessary to simulate large - scale interacting particle systems, such as pedestrian dynamics, collective animal behavior, and inter - cellular interactions. These particle systems can be described by coupled stochastic differential equations (SDEs). When the number of particles tends to infinity, the dynamics of these systems can be described by the McKean - Vlasov equation, which is a nonlinear partial differential equation (PDE). However, it is usually very difficult to directly solve such high - dimensional PDEs because the computational cost increases exponentially with the increase of the state - space dimension.
### Paper Objectives
The main objective of the paper is to develop efficient methods to approximately estimate the function values depending on particle systems under the mean - field limit. Specifically, the author hopes to minimize the computational workload given an error tolerance \( \text{TOL} \). To this end, the author explores different variants of the Monte Carlo method, especially the Multilevel Monte Carlo (MLMC) method and the Multi - index Monte Carlo (MIMC) method.
### Main Contributions
1. **Multilevel Monte Carlo Method (MLMC)**
- The author shows that in the best - case scenario, the work complexity of the MLMC method is \( O(\text{TOL}^{-3}) \).
- By introducing control variates, the MLMC method can significantly reduce the variance, thereby reducing the required number of samples.
2. **Multi - index Monte Carlo Method (MIMC)**
- The author further shows that when using the MIMC method, the work complexity can be further reduced to \( O(\text{TOL}^{-2} \log(\text{TOL}^{-1})^2) \).
- The MIMC method further improves computational efficiency by simultaneously considering multiple discrete parameters (such as the number of time steps and the number of particles).
### Numerical Experiments
The author conducted numerical experiments on the Kuramoto model, which is a coupled oscillator system, to verify the effectiveness of the proposed methods. The experimental results show that the MIMC method indeed has a lower work complexity than the traditional Monte Carlo method and the multilevel Monte Carlo method under the same error tolerance.
### Conclusions
This paper provides an effective means to approximately estimate the function values depending on large - scale particle systems under the mean - field limit by introducing the multilevel Monte Carlo and multi - index Monte Carlo methods. These methods perform well in dealing with high - dimensional problems and can significantly reduce the computational cost, thus having important significance in practical applications.