Improved physics-informed neural network in mitigating gradient related failures

Pancheng Niu, Yongming Chen, Jun Guo, Yuqian Zhou, Minfu Feng +1 more
7/28/2024
cs.LG

Official code: PanChengN/I-PINN (13 stars, last commit 2025-05-21).

Abstract

Physics-informed neural networks (PINNs) integrate fundamental physical principles with advanced data-driven techniques, driving significant advancements in scientific computing. However, PINNs face persistent challenges with stiffness in gradient flow, which limits their predictive capabilities. This paper presents an improved PINN (I-PINN) to mitigate gradient-related failures. The core of I-PINN is to combine the respective strengths of neural networks with an improved architecture and adaptive weights containingupper bounds. The capability to enhance accuracy by at least one order of magnitude and accelerate convergence, without introducing extra computational complexity relative to the baseline model, is achieved by I-PINN. Numerical experiments with a variety of benchmarks illustrate the improved accuracy and generalization of I-PINN. The supporting data and code are accessible at https://github.com/PanChengN/I-PINN.git, enabling broader research engagement.

View on arXivView PDF

Code Implementations(1)

PanChengN/I-PINNOfficial100%
133PythonJul 18, 20241 years ago

Similar Papers

Architecture-Optimization Co-Design for Physics-Informed Neural Networks Via Attentive Representations and Conflict-Resolved Gradients

86% match2026

GradINN: Gradient Informed Neural Network

86% match2024

Implicit Stochastic Gradient Descent for Training Physics-informed Neural Networks

86% match2023

Ensemble learning for Physics Informed Neural Networks: a Gradient Boosting approach

85% match2023

Enhancing Stability of Physics-Informed Neural Network Training Through Saddle-Point Reformulation

85% match2025

LNN-PINN: A Unified Physics-Only Training Framework with Liquid Residual Blocks

85% match2025

Cite this paper

@article{niu2024improved,
  title  = {Improved physics-informed neural network in mitigating gradient related failures},
  author = {Pancheng Niu and Yongming Chen and Jun Guo and Yuqian Zhou and Minfu Feng and Yanchao Shi},
  year   = {2024},
  eprint = {2407.19421},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2407.19421v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2407.19421.svg)](https://codeofpaper.com/paper/2407.19421)

Discussion