graph_classification.pdf

发布时间：2022-05-31 发布人：admin 分类：说明书资料大小：0.72M 资料格式：pdf 举报版权申诉

hehesakura-10305081-4744302542999943284.pdf-第1页.png

第1页 / 共33页

hehesakura-10305081-4744302542999943284.pdf-第2页.png

第2页 / 共33页

hehesakura-10305081-4744302542999943284.pdf-第3页.png

第3页 / 共33页

hehesakura-10305081-4744302542999943284.pdf-第4页.png

第4页 / 共33页

hehesakura-10305081-4744302542999943284.pdf-第5页.png

第5页 / 共33页

hehesakura-10305081-4744302542999943284.pdf-第6页.png

第6页 / 共33页

hehesakura-10305081-4744302542999943284.pdf-第7页.png

第7页 / 共33页

hehesakura-10305081-4744302542999943284.pdf-第8页.png

第8页 / 共33页

文本预览

Graph Classification

Classification Outline • Introduction, Overview • Classification using Graphs – Graph classification – Direct Product Kernel • Predictive Toxicology example dataset – Vertex classification – Laplacian Kernel • WEBKB example dataset • Related Works

Example: Molecular Structures Known Toxic Non-toxic BB AA CC DD CC BB AA DD EE Task: predict whether molecules are toxic, given set of known examples Unknown BB CC AA EE CC BB AA DD DD EE FF

Solution: Machine Learning • Computationally discover and/or predict properties of interest of a set of data • Two Flavors: – Unsupervised: discover discriminating properties among groups of data (Example: Clustering) Data – Property Discovery, Partitioning Clusters – Supervised: known properties, categorize data with unknown properties (Example: Classification) Training Data Build Classification Model Predict Test Data

Classification • Classification: The task of assigning class labels in a discrete class label set Y to input instances in an input space X • Ex: Y = { toxic, non-toxic }, X = {valid molecular structures} Misclassified data instance (test error) Unclassified data instances Training the classification model using the training data Assignment of the unknown (test) data to appropriate class labels using the model

Classification Outline • Introduction, Overview • Classification using Graphs, – Graph classification – Direct Product Kernel • Predictive Toxicology example dataset – Vertex classification – Laplacian Kernel • WEBKB example dataset • Related Works

Classification with Graph Structures • Graph classification • Vertex classification (between-graph) – Each full graph is assigned a class label • Example: Molecular graphs (within-graph) – Within a single graph, each vertex is assigned a class label • Example: Webpage (vertex) / hyperlink (edge) graphs BB AA EE CC Toxic DD NCSU domain Faculty Course Student

Relating Graph Structures to Classes? • Frequent Subgraph Mining (Chapter 7) – Associate frequently occurring subgraphs with classes • Anomaly Detection (Chapter 11) – Associate anomalous graph features with classes • *Kernel-based methods (Chapter 4) – Devise kernel function capturing graph similarity, use vector- based classification via the kernel trick

分享到：

赞收藏

资料库

graph_classification.pdf

相关推荐

人工智能

热门标签

最新资料