Exponentially accurate Spectral Monte Carlo Method for linear PDEs and their error estimates
Jiaying Feng, Changtao Sheng †,⋆, Chenglong Xu †,⋆ · 2025
Read Exponentially accurate Spectral Monte Carlo Method for linear PDEs and their error estimates by Jiaying Feng, Changtao Sheng †,⋆, Chenglong Xu †,⋆ (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-15123.txt.
Introduction
Some equations of physics and finance involve jumps rather than smooth change, driven by processes with heavy-tailed steps. Jiaying Feng, Changtao Sheng, and Chenglong Xu present a numerical method in 2025 for linear equations of that kind, covering both steady and time-changing cases. Their spectral Monte Carlo method blends random sampling with smooth basis functions shaped for the problem.
Control tricks cut the sampling noise, and a walk-on-sphere device handles the geometry. The team proves the error shrinks exponentially fast with each iteration round, for both the steady equation and the time-based one under one unified method. Tests confirm the speed and accuracy in practice. The work extends a method first built for the smoother case of ordinary diffusion.
Catalog record
| CREATOR | Jiaying Feng, Changtao Sheng †,⋆, Chenglong Xu †,⋆ |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-15123
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Exponentially accurate Spectral Monte Carlo Method for linear PDEs and their error estimates Thanks: Submitted to the editors DATE. Jiaying Feng, Changtao Sheng †,⋆, Chenglong Xu †,⋆ Abstract This paper introduces a spectral Monte Carlo iterative method (SMC) for solving linear Poisson and parabolic equations driven by α \alpha -stable Lévy process with α ∈ ( 0 , 2 ) \alpha\in(0,2) , which was initially proposed and developed by Gobet and Maire in their pioneering works (Monte Carlo Methods Appl 10(3-4), 275–285, 2004, and SIAM J Numer Anal 43(3), 1256–1275, 2005) for the case α = 2 \alpha=2 . The novel method effectively integrates multiple computational techniques, including the interpolation based on generalized Jacobi functions (GJFs), space-time spectral methods, control variates techniques, and a novel walk-on-sphere method (WOS). The exponential convergence of the error bounds is rigorously established through finite iterations for both Poisson and parabolic equations involving the integral fractional Laplacian operator. Remarkably, the proposed space-time spectral Monte Carlo method (ST-SMC) for the parabolic equation is unified for both α ∈ ( 0 , 2 ) \alpha\in(0,2) and α = 2 \alpha=2 . Extensive numerical results are provided to demonstrate the spectral accuracy and efficiency of the proposed method, thereby validating the theoretical findings. keywords Spectral method, Monte Carlo method, \alpha -stable L \acute{\text{e}} vy process, error estimate, generalized Jacobi function Funding. The research of the second author is partially supported by the National Natural Science Foundation of China (Nos. 12201385 and 12271365) and the Fundamental Research Funds for the Central Universities 2021110474. The research of the third author is partially supported by the Fundamental Research Funds for the Central Universities. ⋆ Corresponding author. † † runningheads: Spectral Monte Carlo Method for linear PDEs / J. Feng, C. Sheng, and C. Xu AMS 65N35, 65C05, 60G52, 65M15, 33C45 1 Introduction Spectral methods have gained significant popularity in computational fluid dynamics, quantum mechanics, stochastic differential equations, uncertainty quantification, and more (see, e.g., [ 11 , 38 , 40 , 50 , 34 , 60 ] ), primarily due to their ability to provide highly accurate approximations for the solutions of PDEs, as long as these solutions are sufficiently smooth. This is one of the main appealing advantages of spectral methods, the so-called “spectral accuracy”, which is also the goal at the algorithmic level in this paper. Stochastic simulation plays a crucial role in understanding complex real-world phenomena, such as options pricing, turbulence, multiphase flows, molecular dynamics simulations, crystal plasticity, radiation transport, radiotherapy, uncertainty quantification and more [ 10 , 46 , 38 , 7 , 24 , 31 , 37 , 61 , 23 ] , because it encompasses various advantages not typically found in deterministic methods, including being well-suited for solving high-dimensional problems, handling PDEs on complex geometric domains, facilitating parallel computation, and being simpler to implement. The theoretical basis for stochastic methods in solving PDEs resides in the exploitation of a probabilistic representation, namely, the renowned Feynman-Kac formula, which expresses the solution of PDEs as an expectation form. With this, various sampling-based stochastic methods such as the binomial tree method, the multi-tree method, etc., can obtain numerical solutions by directly simulating stochastic processes [ 5 , 4 , 24 , 41 , 48 , 49 ] . Unlike direct simulation methods, the walk-on-spheres method employs transition probabilities to determine the next position of the sphere, thereby simulating stochastic processes on the sphere instead of complete trajectories [ 45 , 17 , 36 , 44 , 62 , 18 , 19 ] . This approach notably reduces computational costs and memory demands and offers an efficient and flexible approach. However, these stochastic methods usually achieve only half-order accuracy, i.e., \mathcal{O}(N^{-1/2}) , where N denotes the number of samples. Although it is possible to improve the accuracy to first-order by making specific modifications, achieving higher-order accuracy becomes extremely challenging. Many efforts have been made to improve the accuracy of stochastic methods for PDEs based on solving backward stochastic differential equations, for instance, the multistep method [ 12 , 28 , 30 , 64 , 35 , 9 ] , the adapted control variates [ 2 , 22 , 25 , 56 ] and so on. Besides, various kinds of higher-order numerical methods, normally second order, have been developed to solve the stochastic PDEs, where high-order typically refers to methods beyond first-order accuracy (cf. [ 3 , 59 , 63 , 21 ] ). However, it is not trivial to directly extend these high-order techniques for solving stochastic PDEs to the approach of simulating the path of a stochastic process that is of interest to this paper. Gobet and Maire [ 26 ] introduced for the first time a spectral Monte Carlo method for the Poisson equation, which combined the idea of iterative variance reduction techniques from sequential Monte Carlo methods with the spectral collocation methods to achieve the Monte Carlo method with spectral accuracy. Subsequently, they updated the proposed method from the perspective of the sequential control variates algorithm, extended it to linear parabolic equations, and proved that the bias and variance of the proposed method decrease geometrically with respect to the number of algorithm steps (cf. [ 27 ] ). The key aspect of those methods lies in the precise reconstruction of the numerical solution from rough data during the iterative process, as well as the terms with different order derivatives of numerical solutions in the governing equation. This accurate reconstruction is achieved by using the connection relationship between derivatives of orthogonal basis functions and their coefficients, thus enabling the SMC method to obtain appealing spectral accuracy. Later, these methods or other similar approaches were further extended to various problems (see, e.g., [ 25 , 28 , 42 , 6 ] and the references therein). We also remark that various techniques for improving sampling efficiency in Monte Carlo methods have been discussed in some recent works, interested readers may refer to cubic stratification [ 16 ] and the Quasi-Monte Carlo method [ 39 ] . On the other hand, some recent studies on stochastic algorithms for PDEs with integral fractional Laplacian operator (IFL), where the IFL with index \alpha\in(0,2) is defined via the hypersingular integral representation (cf. [ 47 ] ): (-\Delta)^{\frac{\alpha}{2}}u(\bm{x})=C_{d,\alpha}\,{\rm p.v.}\!\int_{\mathbb{R}^{d}}\frac{u(\bm{x})-u(\bm{y})}{|\bm{x}-\bm{y}|^{d+\alpha}}\,{\rm d}\bm{y},\quad C_{d,\alpha}:=\frac{2^{\alpha}\Gamma(\frac{d+\alpha}{2})}{\pi^{\frac{d}{2}}\Gamma(1-\frac{\alpha}{2})}, (1) with “p.v.” stands for the principal value and C_{d,\alpha} is the normalization constant. Note that when \alpha=2 the operator coincides with the usual negative Laplacian -\Delta . The non-local and singular nature of IFL poses major difficulties in discretization and analysis for the deterministic numerical methods (cf. [ 1 , 20 , 32 , 57 ] and the references therein). On the contrary, stochastic methods have demonstrated remarkable success in numerical solving this problem, especially in the case of high-dimensions and complex domains. Kyprianou et al. [ 36 ] proposed an unbiased walk-on-sphere method, based on the Feynman-Kac formula, for solving the fractional Poisson equation in two dimensions. Jiao et al. [ 33 ] proposed a modified version of the walk-on-sphere method for solving fractional Poisson equations in high-dimensional balls. Sheng et al. [ 52 ] recently developed an efficient Monte Carlo method for solving fractional PDEs on irregular domains in high dimensions, and also conducted a rigorous error analysis for the proposed