What question did this study set out to answer?

The aim is to enhance the performance of physics-informed neural networks in solving partial differential equations by addressing convergence conflicts.

April 19, 2026MathematicsOpen Access

PDGM-PINN: Partial Derivative Guided Multi-Branch Physics-Informed Neural Network

View Full Paper

Ask AI

Bookmark

Authors

SLShangpeng LeiAl-Farabi Kazakh National University CYChenghan YangAl-Farabi Kazakh National University RGRoberts GrantsRiga Technical University

Discussion

Member takes

Implication

Demonstrates improved performance of a new neural network approach on partial differential equations, indicating efficacy in model complexity reduction.

Key Points

The aim is to enhance the performance of physics-informed neural networks in solving partial differential equations by addressing convergence conflicts.
Developed PDGM-PINN framework to predict both solutions and partial derivatives as dependent variables.
Implemented multi-branch network with sub-branches dedicated to predicting partial derivatives and a main branch for PDE solutions.
Redesigned the loss function to include terms for continuity and consistency of partial derivatives using Schwarz's theorem and Kullback–Leibler divergence.
Conducted experimentation on seven different PDEs alongside ablation, sensitivity, and complexity analyses.
PDGM-PINN outperformed all tested PINN variants in terms of accuracy.
Achieved this with the fewest trainable parameters, highlighting effectiveness in avoiding architectural redundancy.
Demonstrated capability in addressing convergence conflicts in low-dimensional PDEs.

Cite This Study

Lei et al. (2026) studied this question.

synapsesocial.com/papers/69e472d8010ef96374d8ebb8 https://doi.org/10.3390/math14081349

View Full Paper

Ask AI

Bookmark

Implication

Demonstrates improved performance of a new neural network approach on partial differential equations, indicating efficacy in model complexity reduction.

PDGM-PINN: Partial Derivative Guided Multi-Branch Physics-Informed Neural Network

Discussion

Member takes

Implication

Key Points

Cite This Study

PDGM-PINN: Partial Derivative Guided Multi-Branch Physics-Informed Neural Network

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider