PageRank, ProPPR, and Stochastic Logic Programs

摘要

A key feature of ProPPR, a recent probabilistic logic language inspired by stochastic logic programs (SLPs), is its use of personalized PageRank for efficient inference. We adopt this view of probabilistic inference as a random walk over a graph constructed from a labeled logic program to investigate the relationship between these two languages, showing that the differences in semantics rule out direct, generally applicable translations between them.