CS224W摘要01.Introduction； Machine Learning for Graphs

CS224W: Machine Learning with Graphs
Jure Leskovec
2021版

Why Graphs?

Graphs are a general language for describing and analyzing entities with relations/interactions
常见的图数据：
Event Graphs
Computer Networks
Disease Pathways
Food Webs（食物链）
Particle Networks
Underground Networks
Social Networks
Economic Networks
Communication Networks
Citation Networks
Internet
Networks of Neurons
Knowledge Graphs
Gene regulatory network基因调控网络
Scene Graphs从真实图片中分割对象并形成网络
Molecules
Code Graphs
3D Shapes
网络和图的概念上没有很明显的区分

这里给出为什么要使用图模型进行表征：
Complex domains have a rich relational structure, which can be represented as a relational graph.
By explicitly modeling relationships we achieve better performance!

Applications of Graph ML

类型	任务	实例
Node level	Node classification: Predict a property of a node	Categorize online users / items
Edge-level	Link prediction: Predict whether there are missing links between two nodes	Knowledge graph completion
Community(subgraph) level	Clustering: Detect if nodes form a community	Social circle detection
Graph-level	Graph classification: Categorize different graphs	Molecule property prediction
其他	Graph generation、Graph evolution	Drug discovery、Physical simulation

Choice of Graph Representation

Component	Name	Notation
Objects	nodes, vertices	$N$
Interactions	links, edges	$E$
System	network, graph	$G(N,E)$

选择图表征的原则：

类型	特点
professional network	connect individuals that work with each other
sexual networks	connect those that have a sexual relationship
citation network	connect scientific papers that cite each other （有向）

下图来自：https://www.researchgate.net/figure/A-Figure-from-Bearman-Moody-and-Stovel-47-based-the-Add-Health-Data-Set-A-Link_fig2_331307250，是sexual networks，蓝色男性，粉色女性。

除此之外还可以根据图的类型来选择相应的图，有向或无向，带权或不带权，联通或者不联通等。