Abstract:In this paper, we investigate the application of operator learning, specifically DeepONet, to solve partial differential equations (PDEs). Unlike function learning methods that require training separate neural networks for each PDE, operator learning generalizes across different PDEs without retraining. We focus on the performance of DeepONet in Sobolev training, addressing two key questions: the approximation ability of deep branch and trunk networks, and the generalization error in Sobolev norms. Our findings highlight that deep branch networks offer significant performance benefits, while trunk networks are best kept simple. Moreover, standard sampling methods without adding derivative information in the encoding part are sufficient for minimizing generalization error in Sobolev training, based on generalization analysis. This paper fills a theoretical gap by providing error estimations for a wide range of physics-informed machine learning models and applications.

What problem does this paper attempt to address?

The main problem this paper attempts to address is the analysis of generalization ability when solving partial differential equations (PDEs) using deep learning methods, particularly DeepONet. Specifically, the paper focuses on two key issues: 1. **Approximation ability of deep branch and trunk networks**: Investigating the approximation ability of deep branch and trunk networks within deep neural network structures, and exploring whether these networks can provide better approximation errors as their depth increases. 2. **Generalization error in Sobolev training**: Analyzing the generalization error of DeepONet under the Sobolev norm and exploring how to minimize this error by designing network structures. ### Main Contributions - **Approximation ability analysis**: The paper finds that deep branch networks perform better as their depth increases, while trunk networks should maintain a simple structure. This indicates that deep branch networks can benefit from deep structures, whereas trunk networks do not need to be complex. - **Generalization error analysis**: The paper provides an estimate of the generalization error of DeepONet in Sobolev training and points out that standard sampling methods remain effective in Sobolev training without the need to add derivative information in the encoding part. This finding helps simplify network design and avoid unnecessary complexity. ### Research Background Traditional function learning methods (such as Physics-Informed Neural Networks, PINNs) require training a separate neural network for each PDE, which limits their generalization ability. In contrast, operator learning methods (such as DeepONet) can generalize across different PDEs without retraining. Therefore, DeepONet has higher efficiency and generalization ability in solving PDEs. ### Method Overview - **DeepONet structure**: DeepONet consists of branch and trunk networks, where the branch network processes input functions and the trunk network processes coordinates. The paper explores the performance of these two networks as their depth increases. - **Sobolev training**: The paper introduces a loss function under the Sobolev norm to evaluate the performance of DeepONet in solving PDEs. Through theoretical analysis and experimental validation, the paper demonstrates the effectiveness of standard sampling methods in Sobolev training. ### Experimental Results - **Advantages of deep branch networks**: Experimental results show that deep branch networks can significantly improve approximation ability as their depth increases, while trunk networks should maintain a simple structure. - **Minimization of generalization error**: The paper theoretically proves that standard sampling methods remain effective in Sobolev training without the need to add derivative information. This finding simplifies network design and improves the model's generalization ability. ### Conclusion The paper fills the theoretical gap in error estimation under the Sobolev norm for physics-informed machine learning models and provides theoretical support for designing more efficient DeepONet. By optimizing network structures, the performance and generalization ability of DeepONet in solving PDEs can be significantly improved.

DeepONet for Solving PDEs: Generalization Analysis in Sobolev Training

On the training and generalization of deep operator networks

Approximation and Generalization of DeepONets for Learning Operators Arising from a Class of Singularly Perturbed Problems

R-adaptive DeepONet: Learning Solution Operators for PDEs with Discontinuous Solutions Using an R-adaptive Strategy

Training a Deep Operator Network As a Surrogate Solver for Two-Dimensional Parabolic-Equation Models.

An Efficient Method for Solving Two-Dimensional Partial Differential Equations with the Deep Operator Network

Separable Operator Networks

Improved generalization with deep neural operators for engineering systems: Path towards digital twin

Deep Operator Learning Lessens the Curse of Dimensionality for PDEs

Exponential Convergence of Deep Operator Networks for Elliptic Partial Differential Equations

Towards Size-Independent Generalization Bounds for Deep Operator Nets

What do physics-informed DeepONets learn? Understanding and improving training for scientific computing applications

Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators

Ensemble and Mixture-of-Experts DeepONets For Operator Learning

MultiAuto-DeepONet: A Multi-resolution Autoencoder DeepONet for Nonlinear Dimension Reduction, Uncertainty Quantification and Operator Learning of Forward and Inverse Stochastic Problems

XI-DeepONet: An operator learning method for elliptic interface problems

Solving Partial Differential Equations in Different Domains by Operator Learning method Based on Boundary Integral Equations

Learning Partial Differential Equations with Deep Parallel Neural Operator

DOSnet as a Non-Black-Box PDE Solver: When Deep Learning Meets Operator Splitting

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