Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation

Zhoutian Shao, Yuanning Cui, Wei Hu

TL;DR

This work tackles inductive knowledge graph reasoning, where unseen entities challenge traditional embedding-based methods. It proposes MStar, a multi-start progressive propagation model that uses a starting-entities selection (SES) module and a highway layer to broaden query-conditioned propagation, followed by a multi-condition GNN and a decoder; training is aided by LinkVerify to filter uninformative samples. Empirical results on standard inductive KG benchmarks show MStar achieving state-of-the-art performance, especially for distant tail entities, with notable gains in MRR and Hits@10 over strong CMPNN baselines. The approach demonstrates the importance of expanding the propagation scope and incorporating query-aware shortcuts, with practical impact for scalable reasoning over evolving knowledge graphs.

Abstract

Knowledge graphs (KGs) are widely acknowledged as incomplete, and new entities are constantly emerging in the real world. Inductive KG reasoning aims to predict missing facts for these new entities. Among existing models, graph neural networks (GNNs) based ones have shown promising performance for this task. However, they are still challenged by inefficient message propagation due to the distance and scalability issues. In this paper, we propose a new inductive KG reasoning model, MStar, by leveraging conditional message passing neural networks (C-MPNNs). Our key insight is to select multiple query-specific starting entities to expand the scope of progressive propagation. To propagate query-related messages to a farther area within limited steps, we subsequently design a highway layer to propagate information toward these selected starting entities. Moreover, we introduce a training strategy called LinkVerify to mitigate the impact of noisy training samples. Experimental results validate that MStar achieves superior performance compared with state-of-the-art models, especially for distant entities.

Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation

TL;DR

This work tackles inductive knowledge graph reasoning, where unseen entities challenge traditional embedding-based methods. It proposes MStar, a multi-start progressive propagation model that uses a starting-entities selection (SES) module and a highway layer to broaden query-conditioned propagation, followed by a multi-condition GNN and a decoder; training is aided by LinkVerify to filter uninformative samples. Empirical results on standard inductive KG benchmarks show MStar achieving state-of-the-art performance, especially for distant tail entities, with notable gains in MRR and Hits@10 over strong CMPNN baselines. The approach demonstrates the importance of expanding the propagation scope and incorporating query-aware shortcuts, with practical impact for scalable reasoning over evolving knowledge graphs.

Abstract

Knowledge graphs (KGs) are widely acknowledged as incomplete, and new entities are constantly emerging in the real world. Inductive KG reasoning aims to predict missing facts for these new entities. Among existing models, graph neural networks (GNNs) based ones have shown promising performance for this task. However, they are still challenged by inefficient message propagation due to the distance and scalability issues. In this paper, we propose a new inductive KG reasoning model, MStar, by leveraging conditional message passing neural networks (C-MPNNs). Our key insight is to select multiple query-specific starting entities to expand the scope of progressive propagation. To propagate query-related messages to a farther area within limited steps, we subsequently design a highway layer to propagate information toward these selected starting entities. Moreover, we introduce a training strategy called LinkVerify to mitigate the impact of noisy training samples. Experimental results validate that MStar achieves superior performance compared with state-of-the-art models, especially for distant entities.
Paper Structure (32 sections, 10 equations, 2 figures, 9 tables)

This paper contains 32 sections, 10 equations, 2 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Knowledge Graph Reasoning
  4. Inductive Knowledge Graph Reasoning
  5. Conditional Message Passing Neural Networks
  6. Skip Connection
  7. Preliminaries
  8. Knowledge Graph
  9. (Inductive) Knowledge Graph Reasoning
  10. Starting Entities in Progressive Propagation
  11. Methodology
  12. Model Architecture Overview
  13. Starting Entities Selection
  14. Pre-Embedded GNN
  15. Selection
  16. ...and 17 more sections

Figures (2)

  • Figure 1: A motivating example of distant target tail entities for predicting (Univ. of California, Berkeley$\rightarrow$also_known_as$\rightarrow$State Univ.). Prefix "U", "S", and "T" represent university, state, and basketball teams, respectively. Prefix "C" represents category-type entities. Different colors and prefixes symbolize distinct entity types.
  • Figure 2: Framework overview of MStar