Abstract:In many real-world scenarios, the underlying random fluctuations are non-Gaussian, particularly in contexts where heavy-tailed data distributions arise. A typical example of such non-Gaussian phenomena calls for Lévy noise, which accommodates jumps and extreme variations. We propose the Random Batch Method for interacting particle systems driven by Lévy noises (RBM-Lévy), which can be viewed as an extension of the original RBM algorithm in [Jin et al. J Compt Phys, 2020]. In our RBM-Lévy algorithm, $N$ particles are randomly grouped into small batches of size $p$, and interactions occur only within each batch for a short time. Then one reshuffles the particles and continues to repeat this shuffle-and-interact process. In other words, by replacing the strong interacting force by the weak interacting force, RBM-Lévy dramatically reduces the computational cost from $O(N^2)$ to $O(pN)$ per time step. Meanwhile, the resulting dynamics converges to the original interacting particle system, even at the appearance of the Lévy jump. We rigorously prove this convergence in Wasserstein distance, assuming either a finite or infinite second moment of the Lévy measure. Some numerical examples are given to verify our convergence rate.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper aims to solve the problems of efficient simulation of interacting particle systems driven by Lévy processes and their strict convergence analysis. Specifically, the article proposes a Random Batch Method (RBM) to address the computational complexity challenges of traditional direct simulation methods and provides convergence guarantees for this method under Lévy noise.
#### Main problem background
1. **Non - Gaussian fluctuations**: Random fluctuations in many real - world scenarios are non - Gaussian, especially in the case of heavy - tailed data distributions. For example, phenomena such as COVID - 19 death tolls and hurricane - related losses usually need to be modeled with Lévy noise because they can accommodate jumps and extreme changes.
2. **Computational complexity**: For pairwise interacting particle systems, the time complexity of directly evaluating the interactions is \(O(N^{2})\), which is infeasible in large - scale systems.
3. **Limitations of existing methods**: Traditional RBM methods are mainly applicable to continuous noise cases (such as Brownian motion). However, in the case of Lévy noise, especially when the Lévy process has pure - jump characteristics, the existing efficient simulation methods are not yet perfect.
#### Proposed solutions
The article proposes a Random Batch Method for interacting particle systems driven by Lévy processes (RBM - Lévy), with the following main features:
- **Batch processing**: Randomly divide \(N\) particles into several small batches, each of size \(p\), and perform interactions only within each batch at each time interval.
- **Reduced computational complexity**: Through the above batch - processing strategy, the computational complexity is reduced from \(O(N^{2})\) to \(O(pN)\) per time step.
- **Convergence proof**: Strictly prove the convergence of the RBM - Lévy algorithm in Wasserstein distance, regardless of whether the Lévy measure has a finite second moment or an infinite second moment.
#### Key contributions
1. **Extended the application range of RBM**: For the first time, the RBM method is applied to systems driven by Lévy processes, solving the problem that existing methods cannot handle non - Gaussian noise.
2. **Strict mathematical proof**: Provide a detailed convergence analysis to ensure the theoretical reliability of the algorithm.
3. **Numerical verification**: Verify the theoretical results through a series of numerical experiments and show the performance of the algorithm under different parameter settings.
In summary, this paper effectively solves the problems of efficient simulation and convergence analysis of interacting particle systems driven by Lévy processes by introducing the RBM - Lévy method, providing new tools and theoretical support for dealing with complex dynamic systems under non - Gaussian noise.