Microsoft: DAT278x From Graph to Knowledge Graph РAlgorithms and Applications Module 3: Graph Representation Learning Representation learning problem Basically, given an input graph we design and extract structure features for nodes and edges such as different centrality scores for nodes, various similarities scores for one pair of nodes on an link. Basically, a hand-crafted feature matrix is created with expensive human and computing effort. What is the issue here? First, we need to pre-design the features according to our domain knowledge and graph mining experience. The quality of those mining task, therefore, largely depend on the hand-crafted features Second, to do so it will require significant human effort as well as, potentially, very expensive computational cost. Ao inv̩s de construir/modelar manualmente (engeneering) com o esfor̤o de um humano, a ideia ̩ aprender automaticamente (learning). To address these two issues, very recently, feature representation learning is pro
