CS224W(完结)
文章平均质量分 95
斯坦福尤里组的课程笔记,2021版
oldmao_2000
人工不智能,机器不学习。
展开
-
CS224W摘要总纲(已完结)
CS224W: Machine Learning with Graphs课程笔记目录原创 2021-10-24 09:33:16 · 1281 阅读 · 1 评论 -
CS224W摘要01.Introduction; Machine Learning for Graphs
文章目录Why Graphs?Applications of Graph MLCS224W: Machine Learning with GraphsJure LeskovecWhy Graphs?Graphs are a general language for describing and analyzing entities with relations/interactions常见的图数据:Event GraphsComputer NetworksDisease Pathways原创 2021-09-03 11:23:30 · 270 阅读 · 0 评论 -
CS224W摘要02.Traditional Methods for Machine Learning in Graphs
文章目录Node-level Tasks and FeaturesNode degreeNode centralityClustering coefficientGraphlets小结CS224W: Machine Learning with Graphs这节主要从以下三个方面Node-level predictionLink-level predictionGraph-level prediction以无向图为例,讲解传统的特征工程咋整。Node-level Tasks and Featur原创 2021-09-12 10:16:20 · 356 阅读 · 0 评论 -
CS224W摘要03.Node Embedding
文章目录Encoder 和Decoder框架Encoder小例子Note随机游走notation特点算法描述负采样优化SGD小结Node2Vec小结Embedding entire graphs新法 Learn Walk EmbeddingCS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式上一节讲的是传统的图机器学习如何进行特征工程的(node-level, edge-level, graph-level features)而这节课的目标是自动原创 2021-09-12 16:35:24 · 467 阅读 · 0 评论 -
CS224W摘要04.Link Analysis: PageRank
文章目录PageRankLinks as votesMatrix Formulation与随机游走的关系How to solve PageRank?Power IterationProblemsExampleRandom Walk with restarts and Personalized PageRankCS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式从矩阵的角度来理解图,从而可以:Determine node importance vi原创 2021-09-15 13:57:33 · 268 阅读 · 0 评论 -
CS224W摘要05.Label Propagation for Node Classification
文章目录CS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式今天学习message passing框架:Relational classificationIterative classificationBelief propagation原创 2021-09-19 22:03:20 · 691 阅读 · 0 评论 -
CS224W摘要06.07.GNN
文章目录复习Basic of Deep Learning(略)Deep Learning for GraphsSetupDescribe aggregation strategies如何训练单层GNN多层GNN防止over-smoothing如果让GNN go deepCS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式复习之前的知识点就不说了,这里把之前的Shallow Encoder的缺点列一下:????(∣????∣)????(|???原创 2021-09-21 17:37:31 · 256 阅读 · 0 评论 -
CS224W摘要08.Applications of Graph Neural Networks
文章目录CS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式原创 2021-09-24 11:24:42 · 642 阅读 · 0 评论 -
CS224W摘要09.Theory of Graph Neural Networks
文章目录How powerful are GNNs?local neighborhood structuresComputational graphInjective FunctionDesigning the Most Powerful Graph Neural Networkinjective multiset functionGIN vs WL graph KernelCS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式这节内容有点散(虚),原创 2021-09-25 14:59:25 · 491 阅读 · 0 评论 -
CS224W摘要10.Knowledge Graph Embeddings
文章目录Heterogeneous Graphs and Relational GCN (RGCN)Heterogeneous GraphsRGCNblock diagonal matricesBasis/Dictionary learningLink PredictionCS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式主要内容:1.介绍 异质图heterogeneous graph的概念和异质图经典baseline:relational G原创 2021-09-27 15:32:27 · 1234 阅读 · 0 评论 -
CS224W摘要11.Reasoning over Knowledge Graphs
文章目录CS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式本节课将介绍知识图谱上的推理任务(就是QA任务)。目标是回答多跳查询multi-hop queries,包括path queries和conjunctive queries。conjunctive合取的,交集的;与;连接的;联合的,连接(着)的; 契合的;逻辑乘法的介绍query2box方法以解决predictive queries问题。...原创 2021-09-30 14:45:09 · 987 阅读 · 2 评论 -
CS224W摘要12.Frequent Subgraph Mining with GNNs
文章目录CS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式原创 2021-10-04 19:39:28 · 1202 阅读 · 0 评论 -
CS224W摘要13.Community Detection in Networks
文章目录Community Detection in NetworksNetwork CommunitesLouvain AlgorithmDetecting Overlapping Communities: BigCLAMCS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式Community Detection in NetworksNetwork CommunitesLouvain AlgorithmDetecting Overlappi原创 2021-10-07 21:39:44 · 1143 阅读 · 0 评论 -
CS224W摘要14.Traditional Generative Models for Graphs
文章目录CS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式原创 2021-10-09 16:17:51 · 764 阅读 · 0 评论 -
CS224W摘要15.Deep Generative Models for Graphs
文章目录Graph Generative ModelsSetup步骤1步骤2Auto-regressive modelsGraphRNN图与序列的映射RNNGraphRNN训练测试训练实例测试例子问题及解决方法效果可视化效果属性统计相似度深度图生成模型的应用(化学)CS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式这节和上节都是讲图的生成模型,上节讲传统方法,这节讲DL方法。本节两个任务:1.Realistic graph generatio原创 2021-10-11 11:20:53 · 917 阅读 · 1 评论 -
CS224W摘要16.Advanced Topics on GNNs
文章目录Limitations of GNNLimitation 1Limitation 2How to fix them错误思路Position-aware GNNsStructure-aware taskPosition-aware task锚点原力解决方案锚点集位置信息的处理Identity-aware GNNsFailure in three level解决思路Identity-aware GNN[^2]ID-GNN-Fast小结GNN的鲁棒性GNN对抗攻击类型直接攻击间接攻击GNN对抗攻击总览数学原创 2021-10-12 19:15:59 · 680 阅读 · 0 评论 -
CS224W摘要17.Scaling Up GNNs
文章目录Scaling Up GNNs to Large Graphs当前图应用的数据量特点SGD for GNN?Naïve full-batch implementation for GNN?Neighbor Sampling: GraphSAGE[^1]基于邻域的随机梯度下降好处与问题邻居采样好处与问题邻居的选择小结Cluster-GCNSimplified GCNCS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式这节内容貌似也很重要,是针原创 2021-10-16 11:01:04 · 946 阅读 · 1 评论 -
CS224W摘要18.Graph Neural Networks in Computational Biology
文章目录CS224W: Machine Learning with Graphs公式输入请参考:在线Latex公式原创 2021-10-23 21:13:27 · 637 阅读 · 2 评论 -
CS224W摘要19.GNNs for Science
文章目录Pre-Training GNN[^1]Recap挑战Pre-Training问题1Naïve 预训练设置与效果Naïve 预训练效果分析预训练的正确做法Node Level Pre-Training(两个无监督任务)Graph Level Pre-Training(一个有监督任务)小结及结果Hyperbolic Graph Embeddings[^2]表征空间的选择双曲空间模型hyperbolic geometry的相关定义双曲版GNN结果Design Space of GNN[^3]相关术语De原创 2021-10-23 21:14:24 · 901 阅读 · 0 评论