Abstract:Knowledge hypergraphs generalize knowledge graphs using hyperedges to connect multiple entities and depict complicated relations. Existing methods either transform hyperedges into an easier-to-handle set of binary relations or view hyperedges as isolated and ignore their adjacencies. Both approaches have information loss and may potentially lead to the creation of sub-optimal models. To fix these issues, we propose the Hyperbolic Hypergraph Neural Network (H2GNN), whose essential component is the hyper-star message passing, a novel scheme motivated by a lossless expansion of hyperedges into hierarchies. It implements a direct embedding that consciously incorporates adjacent entities, hyper-relations, and entity position-aware information. As the name suggests, H2GNN operates in the hyperbolic space, which is more adept at capturing the tree-like hierarchy. We compare H2GNN with 15 baselines on knowledge hypergraphs, and it outperforms state-of-the-art approaches in both node classification and link prediction tasks.

What problem does this paper attempt to address?

This paper attempts to solve the problems of multi - relation representation and modeling in knowledge hypergraphs. Specifically, it aims to overcome the information loss problem existing in current methods when dealing with hyperedges. Traditional methods usually convert hyperedges into binary relations or ignore their adjacent information, which may lead to the generation of sub - optimal models. To solve these problems, the author proposes a new model - Hyperbolic Hypergraph Neural Network (H2GNN). ### Core Problems of the Paper 1. **Information Loss**: Current methods either simplify hyperedges into easily - handled binary relations or regard hyperedges as isolated entities, ignoring their adjacent relations. Both of these methods will lead to information loss and may thus generate sub - optimal models. 2. **Multi - relation Modeling**: Existing hypergraph neural networks mainly focus on a single type of hyper - relation, ignoring the rich and diverse multi - type hyper - relations in knowledge hypergraphs. 3. **Position - Aware Information**: Current methods often ignore the position information of entities in hyperedges. For example, in a flight path, the position order of Beijing, Shanghai and Guangzhou is semantically meaningful. ### Solutions To solve the above problems, H2GNN introduces a new message - passing mechanism - hyper - star message passing, and utilizes hyperbolic space to better capture tree - like hierarchical structures. Specific improvements include: - **Hyper - star Message Passing**: This is a new scheme for lossless expansion of hyperedges into hierarchical structures, directly embedding adjacent entities, hyper - relations and entity position information. - **Hyperbolic Space**: H2GNN operates in hyperbolic space, which is better at capturing tree - like hierarchical structures and helps to better represent the complex relations in knowledge hypergraphs. - **Position - Aware Information**: By introducing position - aware features, ensure that the order information of entities in hyperedges is correctly considered. ### Experimental Results The paper verifies the superiority of H2GNN in node classification and link prediction tasks by comparing with 15 baseline methods. The experimental results show that H2GNN achieves the best performance on multiple datasets, especially in the node classification task, with higher accuracy and lower standard deviation. ### Summary This paper solves the information loss problem in multi - relation modeling in knowledge hypergraphs by introducing H2GNN, and significantly improves the performance of the model through the application of hyper - star message passing and hyperbolic space. These improvements make H2GNN more effective and robust when dealing with complex knowledge hypergraphs.

Hyperbolic Hypergraph Neural Networks for Multi-Relational Knowledge Hypergraph Representation

Hypergraph Learning: Methods and Practices

Knowledge Association with Hyperbolic Knowledge Graph Embeddings

Learning higher-order features for relation prediction in knowledge hypergraph

Knowledge graph completion model based on hyperbolic hierarchical attention network

Path-based Link Prediction on Hyper-relational Knowledge Graph

Residual Enhanced Multi-Hypergraph Neural Network

HGNN$^+$: General Hypergraph Neural Networks

Hypergraph Transformer Neural Networks

HJE: Joint Convolutional Representation Learning for Knowledge Hypergraph Completion

Hyper-node Relational Graph Attention Network for Multi-modal Knowledge Graph Completion

Dynamic Hypergraph Neural Networks

Link Prediction in Knowledge Graphs Based on Hyperbolic Graph Attention Networks

Dynamic relation learning for link prediction in knowledge hypergraphs

Tensorized Hypergraph Neural Networks

Knowledge Graph Embedding for Hyper-Relational Data

A Hypergraph Neural Network Framework for Learning Hyperedge-Dependent Node Embeddings

Hypergraph Convolutional Network for Multi-Hop Knowledge Base Question Answering (Student Abstract)

Knowledge Hypergraphs: Prediction Beyond Binary Relations

Ha-gnn: a novel graph neural network based on hyperbolic attention