Higher-order neurodynamical equation for simplex prediction.

Area of Science:

• Complex Networks Analysis
• Graph Theory
• Computational Neuroscience

Background:

• Higher-order patterns beyond pairwise relations enhance graph-based models.
• Predicting missing simplices in complex networks offers deeper insights.
• Existing models struggle with arbitrary-order simplices and dynamic feature learning.

Purpose of the Study:

• To address limitations in predicting arbitrary-order simplices.
• To develop a framework integrating neural networks and neurodynamics for simplex prediction.
• To enhance understanding of higher-order structures in complex networks.

Main Methods:

• Introduced Higher-order Neurodynamical Equation for Simplex Prediction (HNESP).
• Simulated dynamical coupling of nodes in simplicial complexes.
• Utilized entropy and normalized multivariate mutual information for simplex-level representations.

Main Results:

• HNESP effectively predicts simplices of arbitrary orders.
• The framework learns node-level representations through dynamical coupling.
• Achieved an average improvement of 8.32% in AUC values over state-of-the-art baselines.

Conclusions:

• HNESP offers a robust approach for higher-order simplex prediction.
• The integration of neurodynamics provides insights into neural network learning mechanisms.
• The proposed method advances the analysis of complex network structures.