Research shows temporal graph knowledge can be reused across different representations

A new research paper explores the reuse of knowledge across different representations of temporal graph data, specifically continuous-time dynamic graphs (CTDGs) and temporal knowledge graphs (TKGs). The study demonstrates that predictive structures learned from relational histories can be effectively reused between these two types of graphs, even when they have distinct input interfaces and independent vocabularies. Furthermore, the research shows that fusing independently pretrained TKG and CTDG learners can enhance prediction accuracy on a third dataset without requiring direct entity or relation-semantic alignment. This suggests that external temporal sources, even without direct alignment, can serve as valuable pretraining resources for recommendation and security detection tasks. AI

IMPACT Demonstrates potential for improved knowledge reuse and model fusion in graph-based AI systems.

RANK_REASON This is a research paper published on arXiv detailing novel findings in machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• arXivLabs
• Bibliographic Explorer
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• CTDGs
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv Recommender
• Influence Flower
• Litmaps
• ScienceCast
• scite Smart Citations
• temporal knowledge graph

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →