Advanced Non-Euclidean Graph Neural Networks Research

Innovating Non-Euclidean Graph Neural Networks

We specialize in theoretical modeling, algorithm development, and experimental validation of geometry-aware graph neural networks in hyperbolic and spherical spaces for advanced data analysis.

Three three-dimensional geometric shapes resembling Ethereum logos, each in different colors, are aligned horizontally. The shapes have reflections on the glossy surface below them, and the background is a smooth gradient with a subtle gray tone.

Exceptional innovation and expertise.

"

Geometric GNNs

Innovative models for non-Euclidean graph neural networks development.

A spherical object with a tessellated pattern of triangular tiles, creating a geometric and futuristic appearance. The pattern is symmetrical, with a reflective quality that emphasizes its three-dimensional shape.

Hyperbolic GNN

This model focuses on adaptive curvature learning in hyperbolic space, effectively capturing hierarchical structures for enhanced performance in various graph-based tasks and applications.

A large geodesic sphere with a faceted surface stands prominently in a landscaped park setting. Several trees, including a palm tree, are positioned in the foreground with blue sky visible above.

Spherical GNN

Designed to model cyclic dependencies, this approach utilizes spherical space closure, particularly effective in analyzing complex interactions such as protein relationships in biological datasets.

A large geodesic sphere structure is set against a backdrop of a partly cloudy sky with bright blue hues. The sphere is composed of triangular panels, giving it a futuristic appearance.

Innovative GNN Models