"GraphSAGE论文阅读报告：王云攀，2019.12.6"

GraphSAGE (Graph Sample and Aggregation) is a graph neural network model proposed by Hamilton et al. in their paper. In this method, the authors aim to learn node embeddings in a scalable and efficient manner by sampling and aggregating information from the graph. The key idea behind GraphSAGE is to generate node embeddings by aggregating features from the node's local neighborhood. This allows the model to capture both the structural information of the graph as well as the features of individual nodes. To achieve this, GraphSAGE employs a sampling strategy that selects a fixed number of neighboring nodes to aggregate information from. These sampled neighbors are then used to calculate an aggregation function, such as mean or max pooling, which produces a new representation for the target node. By performing this operation recursively for multiple layers, GraphSAGE is able to capture information from nodes at varying distances in the graph. In the experiments conducted by Hamilton et al., GraphSAGE was evaluated on several real-world datasets and compared against existing graph embedding methods. The results showed that GraphSAGE outperformed other methods in tasks such as node classification and link prediction. Additionally, the authors demonstrated that GraphSAGE could scale to large graphs with millions of nodes and edges, making it a practical and efficient model for real-world applications. Overall, GraphSAGE is a novel approach to learning node embeddings in graphs by sampling and aggregating information from the local neighborhood. The model's performance and scalability make it a promising tool for a wide range of graph-related tasks, and its potential impact on the field of graph neural networks is significant.