Neuro-Relational Programs: Unifying Queries and Neural Computation over Structured Data
Researchers have introduced Neuro-Relational Programs (NRPs), a novel declarative query language designed to unify relational reasoning with neural computation over structured data. NRPs extend Datalog-style rules to incorporate numeric vector embeddings, enabling the interleaving of relational logic and learnable neural components within a single framework. This approach allows NRPs to function as both trainable query plans and relational-structured neural architectures, offering a general method for neural computation on relational databases. AI
IMPACT This framework could enable more sophisticated and integrated AI models for structured data analysis.