Predicate-Conditional Conformalized Answer Sets for Knowledge Graph Embeddings
This addresses the need for more reliable uncertainty quantification in high-stakes applications like medical diagnosis, representing an incremental improvement over existing marginal coverage methods.
The paper tackles the problem of providing stronger uncertainty guarantees for Knowledge Graph Embedding methods by proposing CondKGCP, which achieves predicate-conditional coverage to ensure consistent reliability per query, with empirical results showing it maintains compact prediction sets.
Uncertainty quantification in Knowledge Graph Embedding (KGE) methods is crucial for ensuring the reliability of downstream applications. A recent work applies conformal prediction to KGE methods, providing uncertainty estimates by generating a set of answers that is guaranteed to include the true answer with a predefined confidence level. However, existing methods provide probabilistic guarantees averaged over a reference set of queries and answers (marginal coverage guarantee). In high-stakes applications such as medical diagnosis, a stronger guarantee is often required: the predicted sets must provide consistent coverage per query (conditional coverage guarantee). We propose CondKGCP, a novel method that approximates predicate-conditional coverage guarantees while maintaining compact prediction sets. CondKGCP merges predicates with similar vector representations and augments calibration with rank information. We prove the theoretical guarantees and demonstrate empirical effectiveness of CondKGCP by comprehensive evaluations.