
刘知远-图神经网络简介.pdf
5星
- 浏览量: 0
- 大小:None
- 文件类型:None
简介:
本文档《刘知远-图神经网络简介》由作者刘知远编写,主要内容为介绍图神经网络的基本概念、发展历程及其在机器学习领域中的应用前景。适合对图数据处理和深度学习感兴趣的读者阅读。
Graphs are valuable data structures in complex real-life applications such as modeling physical systems, learning molecular fingerprints, controlling traffic networks, and recommending friends on social networks. However, these tasks require handling non-Euclidean graph data that contains rich relational information between elements and cannot be effectively managed by traditional deep learning models like convolutional neural networks (CNNs) or recurrent neural networks (RNNs). Nodes in graphs typically contain useful feature information which is not well addressed by most unsupervised representation learning methods. Graph Neural Networks (GNNs) are designed to integrate the feature information and graph structure, enabling better representations on graphs through feature propagation and aggregation. Due to its convincing performance and high interpretability, GNN has recently become a widely used tool for analyzing graphs.
全部评论 (0)


