EPINN: Enhanced Physics-Informed Neural Network for Solving Continuous Integral Equations

2025-12-14 · Source: Journal of Artificial Intelligence Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Scientific Computing · Depth: Expert, quick

Summary

The enhanced physics-informed neural network (EPINN) is a new framework designed to solve high-dimensional and nonlinear integral equations, addressing limitations of traditional numerical methods and existing physics-informed neural networks (PINNs). EPINN integrates three innovations: a variable-order operator decomposition theory to mitigate error accumulation, a differentiable primal function projection layer for physical consistency in Sobolev spaces, and a boundary-aware multi-objective training paradigm for improved generalization. Validated across five benchmark cases in two to four dimensions, including linear/nonlinear Volterra/Fredholm and hybrid integral equations, EPINN reduces relative errors by 1 to 2 orders of magnitude compared to traditional methods, achieving over 92% accuracy with limited training data. It also offers a 3 to 6 times speedup and 23% to 85% error reduction over other deep learning solvers.

Key takeaway

For AI Researchers developing solvers for complex mathematical problems, EPINN demonstrates that integrating specific physical principles into neural networks can yield substantial improvements in accuracy and efficiency. You should consider adopting variable-order operator decomposition and differentiable projection layers to enhance your models' robustness and scalability, particularly for high-dimensional integral equations.

Key insights

EPINN offers a robust, scalable solution for high-dimensional integral equations by integrating physical principles into neural networks.

Principles

• Transform integral equations into well-posed differential systems.
• Ensure physical consistency via primal function projection.
• Improve generalization with boundary-aware multi-objective training.

Method

EPINN employs variable-order operator decomposition, a differentiable primal function projection layer, and a boundary-aware multi-objective training paradigm to solve complex integral equations.

In practice

• Apply EPINN to high-dimensional Volterra/Fredholm equations.
• Utilize EPINN for computational physics problems.
• Integrate physical principles into neural network designs.

Topics

• Enhanced Physics-Informed Neural Networks
• Integral Equations
• High-Dimensional Solvers
• Operator Decomposition Theory
• Multi-objective Training

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• Learning Physical Operators using Neural Operators — A physics-informed framework improves neural operator generalization and temporal continuity for PDE solving via operator splitting.
• Petrov-Galerkin Variational Physics-Informed Neural Network Framework for Two-Dimensional Singularly Perturbed Problems — Petrov-Galerkin VPINN accurately solves 2D singularly perturbed problems by combining neural networks with a variational formulation.
• Extended pseudo-spectral physics-informed neural networks for phase-field models — ESPINN accurately infers phase-field model parameters and free-energy structure from limited transient data, even with noise.
• Error-Conditioned Neural Solvers — Error-conditioned Neural Solvers learn to iteratively correct predictions by directly using PDE residual fields as input.
• Differential Equations V — Physics Informed Neural Networks (PINNs) — Physics Informed Neural Networks embed physical laws into neural networks for fast, accurate, and physically consistent simulations.
• Ten Years of Trying to Learn from a Mesh That’s Too Large — Large CFD meshes pose representational challenges for surrogate models, requiring specialized architectures to capture localized physical features.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Journal of Artificial Intelligence Research.