Stanford CS224W: Machine Learning with Graphs - Stanford CS224W: ML with Graphs | 2021 | Lecture 2.1 - Traditional Feature-based Methods: Node

For more information about Stanford’s Artificial Intelligence professional and graduate programs, visit: https://stanford.io/2ZnSo2T Traditional Feature-based Methods: Node-level features Jure Leskovec Computer Science, PhD Starting from this video, we’ll be discussing the techniques on traditional graph machine learning, especially how to extract features at different levels of graphs. In this video, we talk about node level features and their applications. Node level features focus on characteristics of nodes in the graphs, and can be categorized into importance-based and structure-based ones. To follow along with the course schedule and syllabus, visit: http://web.stanford.edu/class/cs224w/ 0:00 Introduction 0:14 Machine Learning Tasks: Review 0:44 Traditional ML Pipeline 3:02 This Lecture: Feature Design 4:00 Machine Learning in Graphs 4:58 Node-level Tasks 6:01 Node-Level Features: Overview 6:53 Node Features: Node Degree 7:56 Node Features: Node Centrality 8:59 Node Centrality (1) 14:54 Node Features: Clustering Coefficient 17:19 Node Features: Graphlets 25:44 Node-Level Feature: Summary