Keynote: A* Net - A Scalable Path Reasoning Approach for Knowledge Graphs

I’ve always been interested in blending traditional algorithms with deep learning techniques. This keynote comes from a presentation I gave at a knowledge graph seminar.

Inspired by the A* shortest path algorithm, this paper proposes a novel approach to addressing the scalability issues of path-based knowledge graph reasoning in large-scale knowledge graphs. By leveraging a heuristic function trained through deep learning, the proposed method effectively reduces the search space, enabling the selection of critical nodes and edges in each iteration, thereby minimizing both computational time and memory consumption during training and inference.

Here are the slides from my presentation.

Fig.1 Cover Slide

Fig.2 Motivations

Fig.3 Innovations

Fig.4 Preliminary

Fig.5 Preliminary'

Fig.6 Preliminary''

Fig.7 Proposed Method

Fig.8 Proposed Method'

Fig.9 Proposed Method''

Fig.10 Proposed Method'''

Fig.11 Proposed Method: Path-based Reasoning with A*Net

Fig.12 Proposed Method: Path-based Reasoning with A*Net'

Fig.13 Proposed Method: Path-based Reasoning with A*Net''

License

This article is licensed under CC BY-NC-SA 4.0, please cite if you intend to reproduce.

Share