Dimensionality-Aware Fuzzy Graph Neural Networks for Nonlinear Dynamical System Modeling and Quantum-Inspired Pattern Recognition

Authors

  • Dr. N.Shanmuga Vadivu Author
  • Nukala Sujata Gupta Author
  • K. Indumathy Author
  • Elangovan Muniyandy Author
  • K Deepthi Reddy Author
  • S. Pandikumar Author
  • Dr. K. Aruna Bhaskar Author
  • R. Naveen kumar Author

Keywords:

Nonlinear dynamical systems, Fuzzy graph neural networks, Dimensionality-aware learning, Dynamic graph modeling, Quantum-inspired pattern recognition, State prediction.

Abstract

The problem of modeling nonlinear dynamical systems is still difficult due to the involvement of dynamic state evolution, transitions and nonlinear relations. Despite its strong performance in deep learning, the conventional neural network suffers from the inability to model relational structures, and the conventional graph neural network fails to explicitly consider fuzzy uncertainty and different relevance of latent features. In this work, a Dimensionality-Aware Fuzzy Graph Neural Network with Quantum-Inspired Pattern Recognition (DA-FGNN QPR) is proposed for a 5D nonlinear dynamical system. Dynamic graphs are used to represent time-varying state interactions, and Low–Moderate–High fuzzy memberships are used to represent the uncertain operating regions. A mechanism based on the importance of the informative latent features is imposed to control the dimensionality, then quantum inspired amplitude-phase encoding is used to achieve complex pattern discrimination. The framework simultaneously functions as a means of predicting the next state and three-class recognition of Stable, Transitional, and Complex/Unstable regimes. The performance is compared with MLP, GNN and Fuzzy GNN using RMSE, MAE, ????2, Accuracy and Macro F1. The computational results show the value of RMSE is 0.043; MAE is 0.032; ????2,  is 0.968; the accuracy is 94.6 % and the Macro F1 is 0.941, which represents better nonlinear modeling and pattern-recognition capability.

Downloads

Download data is not yet available.

Downloads

Published

2026-09-22

How to Cite

Dimensionality-Aware Fuzzy Graph Neural Networks for Nonlinear Dynamical System Modeling and Quantum-Inspired Pattern Recognition . (2026). Minimax Theory and Its Applications, 11(2s), 813-824. https://journalmta.com/index.php/jmta/article/view/455