Thesis investigates uncertainty in knowledge graph embeddings

This thesis explores uncertainty in knowledge graph embedding (KGE) methods, which represent entities and predicates in vector spaces to infer missing knowledge. It addresses three sources of uncertainty: knowledge uncertainty from incomplete or noisy input, algorithmic uncertainty from stochastic training, and predictive uncertainty in model outputs. The work proposes a voting-based framework to mitigate instability from algorithmic uncertainty and adapts conformal prediction to KGE for distribution-free coverage guarantees. It also develops prediction intervals for confidence-scored triples and an embedding-based approach for probabilistic reasoning over statistical ontologies, aiming for reliable and uncertainty-aware KGE. AI

RANK_REASON The item is an academic paper detailing research on uncertainty in knowledge graph embeddings. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• algorithmic uncertainty
• alphaXiv
• CatalyzeX Code Finder for Papers
• Conformal prediction
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv Recommender
• Influence Flower
• knowledge graph
• knowledge graph embedding
• Litmaps
• Predictive uncertainty
• ScienceCast
• scite Smart Citations
• statistical ontologies

• paper
• other

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