Knowledge graphs represent entities and their relationships as structured networks that support machine-learning applications.

The survey by Ji and colleagues defines knowledge graphs as structured representations of entities and relations, then reviews representation learning, knowledge acquisition, temporal graphs, and knowledge-aware applications. In practice, this gives a learning system more than isolated feature values: it supplies typed relational context. Embedding methods encode entities and relations into vector spaces, while path inference and logical-rule reasoning provide additional mechanisms for predicting or validating links.