Abstract:Solving partial differential equations (PDEs) in Euclidean space with closed-form symbolic solutions has long been a dream for mathematicians. Inspired by deep learning, Physics-Informed Neural Networks (PINNs) have shown great promise in numerically solving PDEs. However, since PINNs essentially approximate solutions within the continuous function space, their numerical solutions fall short in both precision and interpretability compared to symbolic solutions. This paper proposes a novel framework: a closed-form \textbf{Sym}bolic framework for \textbf{PDE}s (SymPDE), exploring the use of deep reinforcement learning to directly obtain symbolic solutions for PDEs. SymPDE alleviates the challenges PINNs face in fitting high-frequency and steeply changing functions. To our knowledge, no prior work has implemented this approach. Experiments on solving the Poisson's equation and heat equation in time-independent and spatiotemporal dynamical systems respectively demonstrate that SymPDE can provide accurate closed-form symbolic solutions for various types of PDEs.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of solving partial differential equations (PDEs), especially by finding symbolic solutions in closed form to improve the accuracy and interpretability of the solutions. Specifically, the paper proposes a new framework - **SymPDE** (Closed - form Symbolic Framework), which uses deep reinforcement learning to directly obtain symbolic solutions of PDEs. #### Background and Challenges 1. **Limitations of Traditional Methods**: - **Difficulties in Analytic Solutions**: For many complex PDEs, traditional mathematical methods are difficult to find analytic solutions. - **Insufficiency of Numerical Solutions**: Existing numerical methods such as Physics - Informed Neural Networks (PINNs) can approximate solutions to a certain extent, but they have deficiencies in accuracy and interpretability when dealing with high - frequency and rapidly changing functions. 2. **Problems of Existing Methods**: - **PINNs**: Although it performs well in numerically solving PDEs, because it is essentially an approximate solution, it performs poorly in high - frequency and steep - change regions and lacks interpretability. - **Other Numerical Methods**: Methods such as DeepONet and FNO are helpful in improving accuracy, but they require a large amount of labeled data for training and still face interpretability challenges. #### Proposed New Method To overcome the above problems, the paper proposes the **SymPDE** framework, with the following main features: 1. **Using Deep Reinforcement Learning**: Generate symbolic solutions of PDEs through deep reinforcement learning, avoiding dependence on complex mathematical theories or labeled data. 2. **Improving Interpretability and Extrapolation Ability**: Symbolic solutions can clearly describe the relationship between solutions and variables, and have stronger interpretability and extrapolation ability. 3. **Applicable to Different Types of PDEs**: Whether it is time - independent PDEs (such as Poisson's equation) or PDEs in space - time dynamic systems (such as the heat conduction equation), SymPDE can provide accurate symbolic solutions. #### Experimental Verification The paper verifies the effectiveness of SymPDE through experiments, specifically including: - **Time - independent Systems**: Such as the solution of Poisson's equation, showing SymPDE's ability to find symbolic solutions in high - dimensional derivative spaces. - **Space - time Dynamic Systems**: Such as the solution of the heat conduction equation, showing SymPDE's advantages in dealing with space - time change problems. In conclusion, this paper provides a new perspective and method for solving PDEs by introducing the SymPDE framework, especially making significant progress in improving the accuracy and interpretability of solutions.

Closed-form Symbolic Solutions: A New Perspective on Solving Partial Differential Equations

Solving Partial Differential Equations Using Deep Learning and Physical Constraints

Interpretable Neural PDE Solvers using Symbolic Frameworks

Symbolically Solving Partial Differential Equations using Deep Learning

A PINN Approach to Symbolic Differential Operator Discovery with Sparse Data

Solving Partial Differential Equations with Point Source Based on Physics-Informed Neural Networks

Symbolic Regression for PDEs using Pruned Differentiable Programs

A Spatial-Temporally Adaptive PINN Framework for 3D Bi-Ventricular Electrophysiological Simulations and Parameter Inference

A Deep Learning Framework for Solving Hyperbolic Partial Differential Equations: Part I

A New Method for Solving Nonlinear Partial Differential Equations Based on Liquid Time-Constant Networks

PinnDE: Physics-Informed Neural Networks for Solving Differential Equations

Physics Informed Neural Networks (PINNs) as intelligent computing technique for solving partial differential equations: Limitation and Future prospects

Operator Learning Enhanced Physics-informed Neural Networks for Solving Partial Differential Equations Characterized by Sharp Solutions

Semi-analytic PINN methods for singularly perturbed boundary value problems

DeepXDE: A Deep Learning Library for Solving Differential Equations

Enforcing continuous symmetries in physics-informed neural network for solving forward and inverse problems of partial differential equations

DISCOVER: Deep identification of symbolically concise open-form PDEs via enhanced reinforcement-learning

Physical informed memory networks for solving PDEs: Implementation and Applications

Physics-Informed neural network solver for numerical analysis in geoengineering

Binary structured physics-informed neural networks for solving equations with rapidly changing solutions

A fully differentiable GNN-based PDE Solver: With Applications to Poisson and Navier-Stokes Equations