How to Turn Your Knowledge Graph Embeddings into Generative Models
This work addresses the challenge of making knowledge graph embeddings more versatile for tasks like sampling and constraint integration, which is incremental but impactful for AI applications involving knowledge graphs.
The paper tackles the problem of turning knowledge graph embeddings into generative models by reinterpreting score functions as circuits, enabling exact maximum-likelihood estimation, efficient sampling, and guaranteed satisfaction of logical constraints with little loss in link prediction performance and improved scalability on large graphs.
Some of the most successful knowledge graph embedding (KGE) models for link prediction -- CP, RESCAL, TuckER, ComplEx -- can be interpreted as energy-based models. Under this perspective they are not amenable for exact maximum-likelihood estimation (MLE), sampling and struggle to integrate logical constraints. This work re-interprets the score functions of these KGEs as circuits -- constrained computational graphs allowing efficient marginalisation. Then, we design two recipes to obtain efficient generative circuit models by either restricting their activations to be non-negative or squaring their outputs. Our interpretation comes with little or no loss of performance for link prediction, while the circuits framework unlocks exact learning by MLE, efficient sampling of new triples, and guarantee that logical constraints are satisfied by design. Furthermore, our models scale more gracefully than the original KGEs on graphs with millions of entities.