Deep Learning-Based Numerical Methods for High-Dimensional Parabolic Partial Differential Equations and Backward Stochastic Differential Equations

By W. E, Jiequn Han, Arnulf Jentzen. 2017

932 citations recorded by the source corpus.

We study a new algorithm for solving parabolic partial differential equations (PDEs) and backward stochastic differential equations (BSDEs) in high dimension, which is based on an analogy between the BSDE and reinforcement learning with the gradient of the solution playing the role of the policy function, and the loss function given by the error between the prescribed terminal condition and the solution of the BSDE. The policy function is then approximated by a neural network, as is done in deep reinforcement learning. Numerical results using TensorFlow illustrate the efficiency and accuracy of the studied algorithm for several 100-dimensional nonlinear PDEs from physics and finance such as the Allen–Cahn equation, the Hamilton–Jacobi–Bellman equation, and a nonlinear pricing model for financial derivatives.

Subfields: Electrical and Electronic Engineering, Finance, Statistical and Nonlinear Physics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Electrical and Electronic Engineering

In Finance

In Statistical and Nonlinear Physics

Citations in the corpus

Wireless Network Intelligence at the Edge (2018, methodology): “[168] W.”
Trainability and Accuracy of Artificial Neural Networks: An Interacting Particle System Approach (2018, background): “These successes open exciting possibilities in applied and computational mathematics that are only beginning to be explored [7–9,17,21,30,36].”
A Multiscale Neural Network Based on Hierarchical Matrices (2018, background): “We have developed a multiscale neural network (MNN) architecture for approximating nonlinear mappings, such as those arising from the solution of integral equations (IEs) or partial differential equations (PDEs).”
On universal approximation and error bounds for Fourier Neural Operators (2021, background): “For both these PDEs, we prove rigorously that there exists a Ψ-FNO which can approximate the underlying nonlinear operators efficiently, as we can show that the size of the Ψ-FNO only needs to grow polynomially in terms of the error.”
Deep Autoencoder based Energy Method for the Bending, Vibration, and Buckling Analysis of Kirchhoff Plates (2020, methodology): “numerical method for high-dimensional parabolic PDEs and back-forward stochastic differential equations [38, 39].”
Neural Networks as Interacting Particle Systems: Asymptotic Convexity of the Loss Landscape and Universal Scaling of the Approximation Error (2018, background): “These successes open exciting possibilities in applied and computational mathematics that are only beginning to be explored [5–7,12,16,25,30].”
Forward-Backward Stochastic Neural Networks: Deep Learning of High-dimensional Partial Differential Equations (2018, methodology): “Similar to their prior works [12, 13], the authors then devise an algorithm to compute Y0 = u(0, X0) = u(0, ξ) by treating Y0, Z0 = Du(0, ξ), Γ0 = D(2)u(0, ξ), and A0 = LDu(0, ξ) as parameters of their model.”
Integrating Physics-Based Modeling with Machine Learning: A Survey (2020, neutral): “Solve PDEs [147] [65] [140] [9] [215] [42] [264] [106] [235] [107] [131] [209] [234] [279] [281] [60] [177] [208] [231] [301] [71] [89] [129] [268] [194] [176] [16] [48] [218] [44] [60] [172] [235] [19] [51] [131] [76] [278] [194] [176] [206] [166]”
Machine Learning Approximation Algorithms for High-Dimensional Fully Nonlinear Partial Differential Equations and Second-order Backward Stochastic Differential Equations (2017, methodology): “The derivation of the deep 2BSDE method is mainly based on ideas in E, Han, & Jentzen [33] and Cheridito et al. [22] (cf., e.g., [33, Section 2] and [22, Theorem 4.10]).”
Characterizing possible failure modes in physics-informed neural networks (2021, methodology): “A number of tools and methodologies now exist to solve scientiﬁc problems by combining ML and domain insights [14, 20, 27, 28, 38].”
Solving inverse problems using data-driven models (2019, background): “Elad (2015), Patch-disagreement as a way to improve K-SVD denoising.”
A unified deep artificial neural network approach to partial differential equations in complex geometries (2017, methodology): “In the following we let B in (3.1) be the identity operator and we hence consider PDEs with Dirichlet boundary conditions.”