Abstract:The pursuit of analytical solutions for differential equations has historically been limited by the need for extensive prior knowledge and mathematical prowess; however, machine learning methods like genetic algorithms have recently been applied to this end, albeit with issues of significant time consumption and complexity. This paper presents a novel machine learning-based solver, SSDE (Symbolic Solver for Differential Equations), which employs reinforcement learning to derive symbolic closed-form solutions for various differential equations. Our evaluations on a range of ordinary and partial differential equations demonstrate that SSDE provides superior performance in achieving analytical solutions compared to other machine learning approaches.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to address the challenges encountered in the process of finding analytical solutions to differential equations (DEs). Specifically, the author points out: 1. **Limitations of traditional methods**: - Traditional methods for finding analytical solutions require a great deal of prior knowledge and complex mathematical techniques. - Deep - learning methods such as Physics - Informed Neural Networks (PINNs) can approximate the solution of differential equations, but they have deficiencies in interpretability and satisfying physical constraints. - Symbolic regression methods (such as genetic algorithms) can generate symbolic expressions, but they have high computational complexity and the generated expressions are often too complex to be interpretable. 2. **Problems with existing machine - learning methods**: - Deep - learning methods (such as PINNs) rely on numerical solutions when solving differential equations, and these numerical solutions may have error - propagation problems, resulting in the final symbolic solution being not precise enough. - Symbolic regression methods such as genetic algorithms can generate symbolic solutions, but in high - dimensional problems, the search space is huge and the computational cost is high. - Methods such as Kolmogorov - Arnold Networks (KANs) will have a significant decrease in precision when converting the learned representation into a symbolic expression. 3. **The new method proposed in the paper**: - The author introduces a new framework SSDE (Symbolic Solver for Differential Equations) based on reinforcement learning for directly deriving closed - form symbolic solutions of differential equations. - SSDE improves the efficiency and accuracy of the solution by combining physical constraints and reinforcement learning, and can generate more concise and more interpretable symbolic solutions. - The Risk - Seeking Constant Optimization (RSCO) algorithm is proposed to accelerate convergence and maintain precision. - A recursive exploration technique is developed for solving high - dimensional partial differential equations (HD - PDEs), enabling SSDE to find accurate closed - form solutions in multi - dimensional problems. ### Summary The main objective of the paper is to overcome the limitations of existing methods in finding analytical solutions to differential equations by introducing a new reinforcement - learning framework SSDE, providing a new method that is efficient, accurate, and interpretable.

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

Genetic Programming Based Symbolic Regression for Analytical Solutions to Differential Equations

Symbolically Solving Partial Differential Equations using Deep Learning

General solutions for nonlinear differential equations: a rule-based self-learning approach using deep reinforcement learning

A Neuro-Symbolic Method for Solving Differential and Functional Equations

A Survey on Solving and Discovering Differential Equations Using Deep Neural Networks

Extreme Learning Machine with Kernels for Solving Elliptic Partial Differential Equations

DeepXDE: A Deep Learning Library for Solving Differential Equations

Learning Differential Equations that are Easy to Solve

Solving High-Dimensional Partial Differential Equations Using Deep Learning.

A Normal Equation-Based Extreme Learning Machine for Solving Linear Partial Differential Equations

D-CIPHER: Discovery of Closed-form Partial Differential Equations

Solving Nonlinear Equations Systems with an Enhanced Reinforcement Learning Based Differential Evolution

Solving the 2D Advection-Diffusion Equation using Fixed-Depth Symbolic Regression and Symbolic Differentiation without Expression Trees

Learning To Solve Differential Equation Constrained Optimization Problems

Learning To Solve Differential Equations Across Initial Conditions

A forward differential deep learning-based algorithm for solving high-dimensional nonlinear backward stochastic differential equations

Data-driven discovery of free-form governing differential equations

Symbolic Equation Solving via Reinforcement Learning

Solving Partial Differential Equations Using Deep Learning and Physical Constraints

Local Extreme Learning Machines and Domain Decomposition for Solving Linear and Nonlinear Partial Differential Equations