CS224W: Machine Learning with Graphs | 2021 | Lecture 2.3 - Traditional Feature-based Methods: Graph

For more information about Stanford’s Artificial Intelligence professional and graduate programs, visit: Traditional Feature-based Methods: Graph-level features Jure Leskovec Computer Science, PhD In this video, we focus on extracting features from the graphs as a whole. In other words, we want features that characterize the structure of entire graphs. Specifically, we’re interested in graph kernel methods that measure the similarity between two graphs. We’ll describe different approaches to extracting such graph kernels, including Graphlet features and WL kernels. To follow along with the course schedule and syllabus, visit: 0:00 Introduction 0:56 Background: Kernel Methods 2:48 Graph-Level Features: Overview 3:25 Graph Kernel: Key Idea 5:56 Graphlet Features 8:20 Graphlet Kernel 12:35 Color Refinement (1) 15:13 Weisfeiler-Lehman Graph Features 16:32 Weisfeiler-Lehman Kernel 17:45 Graph-Level Features: Summary 19:07 Today’s Summary