Score-fPINN: Fractional Score-Based Physics-Informed Neural Networks for High-Dimensional Fokker-Planck-Levy Equations

Zheyuan Hu,Zhongqiang Zhang,George Em Karniadakis,Kenji Kawaguchi

2024-06-17

Abstract:We introduce an innovative approach for solving high-dimensional Fokker-Planck-Lévy (FPL) equations in modeling non-Brownian processes across disciplines such as physics, finance, and ecology. We utilize a fractional score function and Physical-informed neural networks (PINN) to lift the curse of dimensionality (CoD) and alleviate numerical overflow from exponentially decaying solutions with dimensions. The introduction of a fractional score function allows us to transform the FPL equation into a second-order partial differential equation without fractional Laplacian and thus can be readily solved with standard physics-informed neural networks (PINNs). We propose two methods to obtain a fractional score function: fractional score matching (FSM) and score-fPINN for fitting the fractional score function. While FSM is more cost-effective, it relies on known conditional distributions. On the other hand, score-fPINN is independent of specific stochastic differential equations (SDEs) but requires evaluating the PINN model's derivatives, which may be more costly. We conduct our experiments on various SDEs and demonstrate numerical stability and effectiveness of our method in dealing with high-dimensional problems, marking a significant advancement in addressing the CoD in FPL equations.

Machine Learning,Dynamical Systems,Numerical Analysis

What problem does this paper attempt to address?

This paper aims to address the problem of high-dimensional Fokker-Planck-Lévy (FPL) equations. FPL equations are used in fields such as physics, finance, and ecology to model non-Brownian processes. However, solving high-dimensional FPL equations poses significant challenges due to the non-locality introduced by fractional derivatives and the "curse of dimensionality" (CoD), where computational requirements grow exponentially with the dimension. To tackle this issue, the authors propose an innovative method—Score-based Physics-Informed Neural Networks (Score-fPINN). This method transforms the FPL equation into a second-order partial differential equation without fractional Laplacian operators by introducing a fractional score function, allowing the use of standard Physics-Informed Neural Networks (PINN) for solving. Additionally, the authors propose two methods for obtaining the fractional score function: Fractional Score Matching (FSM) and Score-fPINN. - **FSM**: Suitable for cases with known conditional distributions, it obtains the fractional score function by minimizing the mean squared error between the fractional score function and the true conditional fractional score function. - **Score-fPINN**: Does not rely on the specific form of the stochastic differential equation (SDE) but requires computing the derivatives of the neural network model with respect to the inputs, making it more costly. Experimental results show that this method demonstrates numerical stability and effectiveness in handling high-dimensional problems, significantly advancing the research progress of high-dimensional FPL equations. In summary, this work aims to overcome the limitations of traditional methods and existing technologies in solving high-dimensional FPL equations.

Score-fPINN: Fractional Score-Based Physics-Informed Neural Networks for High-Dimensional Fokker-Planck-Levy Equations

Score-Based Physics-Informed Neural Networks for High-Dimensional Fokker-Planck Equations

Tackling the Curse of Dimensionality in Fractional and Tempered Fractional PDEs with Physics-Informed Neural Networks

Bi-orthogonal fPINN: A physics-informed neural network method for solving time-dependent stochastic fractional PDEs

Laplace-fPINNs: Laplace-based fractional physics-informed neural networks for solving forward and inverse problems of subdiffusion

Physics-informed Neural Networks for Functional Differential Equations: Cylindrical Approximation and Its Convergence Guarantees

Physics-Informed Solution of The Stationary Fokker-Plank Equation for a Class of Nonlinear Dynamical Systems: An Evaluation Study

fPINNs: Fractional Physics-Informed Neural Networks

Monte Carlo fPINNs: Deep learning method for forward and inverse problems involving high dimensional fractional partial differential equations

Deep fuzzy physics-informed neural networks for forward and inverse PDE problems

A novel discretized physics-informed neural network model applied to the Navier-Stokes equations

ELM-FBPINN: efficient finite-basis physics-informed neural networks

Physics-Informed Kernel Function Neural Networks for Solving Partial Differential Equations

Monte Carlo PINNs: deep learning approach for forward and inverse problems involving high dimensional fractional partial differential equations

Finite basis physics-informed neural networks (FBPINNs): a scalable domain decomposition approach for solving differential equations

A novel optimization-based physics-informed neural network scheme for solving fractional differential equations

A Physics Informed Neural Network for Time-Dependent Nonlinear and Higher Order Partial Differential Equations

GMC-PINNs: A new general Monte Carlo PINNs method for solving fractional partial differential equations on irregular domains

Advanced Physics-Informed Neural Networks for Numerical Approximation of the Coupled Schrödinger-Kdv Equation

FO-PINNs: A First-Order formulation for Physics Informed Neural Networks

Physical informed neural networks with soft and hard boundary constraints for solving advection-diffusion equations using Fourier expansions