CS224W Lecture 1 Introduction

  • Why Graphs?
    • effective data structure to describe and analyze entities with relations.
  • How to we take advantages of the graph for better prediction?
    • Explicitly modeling relationships
  • Modern ML systems are designed for simple sequences and grids
    • but graphs have…
      • arbitrary size and complex topological structure
      • no fixed ordering
  • We can develop neural net that are much more broadly applicable by modeling Graphs
  • Representation learning will be the main topic for this course
    • automatically learn the features without tuning the features each time
  • ML tasks related to graphs: node classification, link prediction, graph classification, clustering, …
  • choice of the proper network representation of a given domain/problem determines our ability to use netowrk successfully
    • the way we assign links will determine the nature of the question we can study.
    • directed or undirected
    • bipartite graph
    • edge list, adjacency matrix or adjecency list
    • node and edge attributes
    • weighted edges, self edges, multigraphs, ..
October 19, 2021
Tags: cs224w