Abstract:In recent years, data-driven methods have been developed to learn dynamical systems and partial differential equations (PDE). The goal of such work is discovering unknown physics and the corresponding equations. However, prior to achieving this goal, major challenges remain to be resolved, including learning PDE under noisy data and limited discrete data. To overcome these challenges, in this work, a deep-learning based data-driven method, called DL-PDE, is developed to discover the governing PDEs of underlying physical processes. The DL-PDE method combines deep learning via neural networks and data-driven discovery of PDE via sparse regressions. In the DL-PDE, a neural network is first trained, and then a large amount of meta-data is generated, and the required derivatives are calculated by automatic differentiation. Finally, the form of PDE is discovered by sparse regression. The proposed method is tested with physical processes, governed by groundwater flow equation, convection-diffusion equation, Burgers equation and Korteweg-de Vries (KdV) equation, for proof-of-concept and applications in real-world engineering settings. The proposed method achieves satisfactory results when data are noisy and limited.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: discovering partial differential equations (PDEs) from discrete and noisy data. Specifically, the author proposes a data - driven method based on deep learning (DL - PDE), aiming to overcome the challenges of existing methods in dealing with limited discrete data and noisy data. Traditional methods are prone to generating unstable numerical derivatives when dealing with noisy data, thus affecting the discovery accuracy of partial differential equations. In addition, existing methods usually require a large amount of data and are sensitive to noise. ### Main problem description of the paper 1. **Treatment of noisy data**: - Traditional numerical methods (such as the finite difference method) may lead to unstable and ill - conditioned results when dealing with noisy data. - A method that can accurately discover PDEs under noisy data conditions is required. 2. **Treatment of limited discrete data**: - In practical applications, only limited discrete observation data can often be obtained. - A method that can accurately discover PDEs using limited data is required. 3. **Limitations of existing methods**: - Existing methods (such as sparse regression, Gaussian process, neural network, etc.) have limitations when dealing with PDEs of unknown structure. - A method that combines multiple advantages is required to improve the accuracy and robustness of PDE discovery. ### Core idea of the DL - PDE method - **Combination of deep neural network and sparse regression**: - Use a deep neural network to fit data and generate meta - data. - Use automatic differentiation to calculate derivatives, avoiding the instability brought by numerical differentiation. - Use the sparse regression method to identify the key terms that constitute PDEs from the candidate library. - **Treatment of noise and limited data**: - Generate a large amount of meta - data to make up for the deficiency of limited observation data. - The automatic differentiation technique makes it more robust to noisy data. ### Experimental verification The paper verifies the effectiveness of the DL - PDE method through four classical physical processes (groundwater flow equation, convection - diffusion equation, Burgers equation, and KdV equation). The experimental results show that even in the case of noisy data and limited data, DL - PDE can accurately discover PDEs with a relatively small error. ### Conclusion The DL - PDE method has successfully solved the challenges of existing methods in dealing with noisy data and limited discrete data, providing an effective method for discovering partial differential equations from complex data.

DL-PDE: Deep-learning based data-driven discovery of partial differential equations from discrete and noisy data

Deep-learning based discovery of partial differential equations in integral form from sparse and noisy data

Deep-learning of parametric partial differential equations from sparse and noisy data

DLGA-PDE: Discovery of PDEs with incomplete candidate library via combination of deep learning and genetic algorithm

PDE-LEARN: Using deep learning to discover partial differential equations from noisy, limited data

Machine Learning of Partial Differential Equations from Noise Data

Discovering Nonlinear PDEs from Scarce Data with Physics-encoded Learning

Data-driven discovery of PDEs in complex datasets

Physics-Guided Discovery of Highly Nonlinear Parametric Partial Differential Equations

PDE-READ: Human-readable partial differential equation discovery using deep learning

Solving Partial Differential Equations Using Deep Learning and Physical Constraints

Data-Driven Deep Learning of Partial Differential Equations in Modal Space

Learning Partial Differential Equations from Data Using Neural Networks

PDE-Net: Learning PDEs from Data

Physics-Based Deep Learning for Flow Problems

Data-driven Recovery of PDE Models and Unveiling of Solution Interconnections

Physics-informed learning of governing equations from scarce data

Numerical solution for high-dimensional partial differential equations based on deep learning with residual learning and data-driven learning

A Deep Learning Approach for Predicting Spatiotemporal Dynamics From Sparsely Observed Data

DISCOVER: Deep Identification of Symbolically Concise Open-Form Partial Differential Equations Via Enhanced Reinforcement Learning

PDE-Net 2.0: Learning PDEs from Data with A Numeric-Symbolic Hybrid Deep Network