Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings

Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第1张图片
Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第2张图片
形式化定义network embedding如上,我们的目标就是让相似的节点拥有更相似的特征。
Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第3张图片当然最直接的衡量节点相似性的方式就是,如果两个节点u,v相连,则认为两个节点是相似的。即
z v T z u = A u , v z_v^Tz_u = A_{u,v} zvTzu=Au,v
Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第4张图片
Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第5张图片
前几节所讲的deep walk以及node2vec是一种通过随机游走的来判断节点相似性的更复杂的方式。
Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第6张图片
Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第7张图片
Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第8张图片

Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第9张图片Machine Learning with Graphs 之 Matrix Factorization and Node Embeddings_第10张图片

你可能感兴趣的:(阅读论文,机器学习,图论)