DeepHGCN: Toward Deeper Hyperbolic Graph Convolutional Networks

Abstract

Hyperbolic graph convolutional networks (HGCNs) have demonstrated significant potential in extracting information from hierarchical graphs. However, existing HGCNs are limited to shallow architectures due to the computational expense of hyperbolic operations and the issue of over-smoothing as depth increases. Although treatments have been applied to alleviate over-smoothing in GCNs, developing a hyperbolic solution presents distinct challenges since operations must be carefully designed to fit the hyperbolic nature. Addressing these challenges, we propose DeepHGCN, the first deep multi-layer HGCN architecture with dramatically improved computational efficiency and substantially reduced over-smoothing. DeepHGCN features two key innovations: (1) a novel hyperbolic feature transformation layer that enables fast and accurate linear mappings, and (2) techniques such as hyperbolic residual connections and regularization for both weights and features, facilitated by an efficient hyperbolic midpoint method. Extensive experiments demonstrate that DeepHGCN achieves significant improvements in link prediction and node classification tasks compared to both Euclidean and shallow hyperbolic GCN variants.

Accepted to IEEE Transactions on Artificial Intelligence

References in corpus (17)

• Adam: A Method for Stochastic Optimization
• A Simple Framework for Contrastive Learning of Visual Representations
• Hierarchical structure and the prediction of missing links in networks
• Simplifying Graph Convolutional Networks
• Hyperbolic Geometry of Complex Networks
• Representation Learning on Graphs with Jumping Knowledge Networks
• Popularity versus Similarity in Growing Networks
• DropEdge: Towards Deep Graph Convolutional Networks on Node Classification
• Predict then Propagate: Graph Neural Networks meet Personalized PageRank
• Hyperbolic Graph Convolutional Neural Networks
• Graph Neural Networks Exponentially Lose Expressive Power for Node Classification
• Riemannian Adaptive Optimization Methods
• HRCF: Enhancing Collaborative Filtering via Hyperbolic Geometric Regularization
• Dirichlet Energy Constrained Learning for Deep Graph Neural Networks
• Fully Hyperbolic Neural Networks
• Power Consumption Variation over Activation Functions
• Towards Scalable Hyperbolic Neural Networks using Taylor Series Approximations