Knowledge Inference for Graphs

This lecture by the instructor covers the concept of knowledge inference for graphs, focusing on propagating attribute values among connected nodes in a graph. It discusses label inference, optimization objectives, injecting pre-existing knowledge, label propagation algorithms, and the behavior of probabilities in the inference process. The lecture also explores extensions of label propagation, the semi-supervised learning aspect, and the application of knowledge inference in completing knowledge bases. Various models and algorithms for knowledge base completion and learning are presented, along with their practical implications and references to related research.