New FICE model offers fully inductive cardinality estimation for knowledge graphs

Researchers have developed FICE, a novel graph neural network designed for fully inductive cardinality estimation in SPARQL queries over knowledge graphs. Unlike previous methods that require retraining for new or modified graphs, FICE generalizes to unseen graphs and relations without re-training. The model utilizes a two-component GNN, with an encoder generating embeddings and a decoder predicting log-cardinality, trained jointly for specialized estimation. FICE demonstrates significant improvements over existing methods, reducing median query error and enhancing tail behavior across multiple knowledge graphs. AI

IMPACT This research could improve the efficiency and scalability of querying large knowledge graphs, potentially impacting AI systems that rely on structured knowledge.

RANK_REASON This is a research paper detailing a new method for query optimization. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Basic Graph Patterns
• graph neural network
• Hugging Face
• Knowledge Graphs
• SPARQL

• paper
• infra

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