Teacher Zhao will not write anything about it
The purpose of learning graph neural networks
The urgent new needs of the deep learning era
·Many actual application scenarios of data are generated from non-Euclidean space. How to apply the deep learning method to graph data
·Difficulty: the irregularity of the graph (disordered nodes, different numbers of neighbors)
·Wide application areas: e-commerce, financial risk control, recommendation systems, social networks, medical molecular prediction
·Academic new hotspots (according to the articles collected by NIPS and KDD in 2020)
List of papers
1. Node2vec: balance homogeneity and structure
2. LINE: 1st order +
2nd order similarity 3. SDNE: Multi-layer
autoencoder 4. Metapath2vec: Heterogeneous graph network
5. TransE: Foundation of knowledge graph, and a series Article
6. GCN: The pioneering work of graph convolution
7. GAT: GNN + attention mechanism
8. MPNN: Spatial convolution messaging framework
9. GGNN: Gated graph neural network
10. GraphSAGE: Inductive learning framework
GNN development
2013 Spectral networks and locally connected networks on graphs
2013 Translating Embeddings for Modeling Multi-relational Data
2014 DeepWalk: Online Learning of Social Representations
2016 Semi-Supervised Classification with Graph Convolutional Networks
Three surveys:
A Comprehensive Survey on Graph Neural Networks
Deep Learning on Graphs: A Survey
Graph Neural Networks: A Review of Methods and Applications
A book:
Introduction to Graph Neural Networks.
Zhiyuan Liu, Jie Zhou,
a new book by Liu Zhiyuan , Tsinghua University, There is no Chinese yet, there is a translation on
Zhihu https://zhuanlan.zhihu.com/p/129305050