Personalized PageRank

Personalized PageRank is a variant of the original PageRank algorithm where the random-walk teleport vector concentrates probability on a chosen seed set instead of distributing it uniformly across the graph. The result is a ranking of every node by its importance relative to those seeds. Nodes connected through many high-weight paths from the seeds rank high, distant or weakly-connected nodes rank low. PPR is the retrieval primitive inside Concord by IaxaI's knowledge graph. A user query embeds, dual-path vector search picks candidate phrase and passage nodes above similarity thresholds, those nodes seed a PPR run, and the resulting node weights blend with vector similarity and recency to produce the final retrieval. PPR captures multi-hop reasoning that pure embedding similarity misses. An alert connects to a prior incident through a shared host, which connects to a regulatory control through an evidence chain. The walk finds the path. The score quantifies it.