GNN中三种基础神经网络：GCN, GAT, GraphSAGE.代表了3种聚合的思想。
这是我的个人理解(不权威)。然随着图神经网络的快速发展，在图神经网络种添加各种组件、搭建新结构或着引入新方法等越来越多，逐渐令人眼花缭乱。特此，在这里整理相关论文，加以梳理，使脉络清晰。提醒：这里的论文大多涉及交通领域。
| 简称 | 文章 | 作用 | 备注 |
|---|---|---|---|
| Diffusion convolutional recurrent neural network: Data-driven traffic forecasting | 扩散图卷积 | ||
| STGCN | Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting | GCN捕捉空间信息，门控捕捉时间信息 | 我的《代码分析文集》 |
| Dynamic spatiotemporal graph convolutional neural networks for traffic data imputation with complex missing patterns | 动态图学习？ | 原文 |
| 简称 | 文章 | 作用 | 备注 |
|---|---|---|---|
| CAFMGCN | Cross-Attention Fusion Based Spatial-Temporal Multi-Graph Convolutional Network for Traffic Flow Prediction | 捕捉3个时间段的数据，捕捉3个不同语义的邻接矩阵,然后使用T-atten和S-atten交叉捕捉信息并融合 | 只有data,No code😢 |
| Joint predictions of multi-modal ride-hailingdemands: A deep multi-task multi-graph learning-based approach | 原文 |
| 简称 | 文章 | 作用 | 备注 |
|---|---|---|---|
| Deep spatial–temporal sequence modeling for multi-step passenger demand prediction | 使用的输入数据为历史需求+气象数据+时间元，预定义矩阵和自适应矩阵搭建的GCN直接合并。采用encoder-decoder结构。decoder是LSTM并用于进行多步预测。 | 我的博文 |
M A E = 1 n ∑ i = 1 n ∣ y i ^ − y i ∣ MAE=\frac{1}{n}\sum_{i=1}^n \lvert \hat{y_i}-y_i\rvert MAE=n1i=1∑n∣yi^−yi∣
R M S E = 1 n ∑ i = 1 n ( y i ^ − y i ) RMSE=\sqrt{\frac{1}{n}\sum_{i=1}^n (\hat{y_i}-y_i)} RMSE=n1i=1∑n(yi^−yi)
M A P E = 1 n ∑ i = 1 n ∣ y i ^ − y i y i ∣ MAPE=\frac{1}{n}\sum_{i=1}^n \lvert \frac{\hat{y_i}-y_i}{y_i}\rvert MAPE=n1i=1∑n∣yiyi^−yi∣
s M A P E = 200 n ∑ i = 1 n ∣ x i − y i x i + y i ∣ sMAPE=\frac{200}{n}\sum_{i=1}^{n}\lvert \frac{x_i-y_i}{x_i+y_i} \rvert sMAPE=n200i=1∑n∣xi+yixi−yi∣
| 论文名称 | 简称 | 损失函数 | 备注 |
|---|---|---|---|
| Deep learning in the COVID-19 epidemic: A deep model for urban traffic revitalization index | DeepTRI | MSE\RMSE\MAPE | 我博 |
| Dynamic Spatiotemporal Graph Convolutional Neural Networks for Traffic Data Imputation with Complex Missing Patterns | DSTGCN | MAE\RMSE | 我博 |
| Adaptive Multi-channel Graph Convolutional | AM-GCN | ACC\F1 | 我博 |
| A deep-learning model for urban traffic flow prediction with traffic events mined from twitter | MAE\RMSE\sMAPE | 我博 | |
| Meta Graph Transformer: A Novel Framework for Spatial–Temporal Traffic Prediction | MGT | RMSE\MAE\MAPE | 我博 |
| Deep spatial–temporal sequence modeling for multi-step passenger demand prediction | RMSE\MAE\MAPE | 我博 | |
| elf-attention based generative adversarial network for traffic flow prediction | SATP-GAN | MAE\MSE\RMSE | 我博 |
| Joint Demand Prediction for Multimodal Systems: A Multi-task Multi-relational Spatiotemporal Graph Neural Network Approach | ST-MRGNN | RMSE\MAE\R^2 | 我博 |
| Cross-Attention Fusion Based Spatial-Temporal Multi-Graph Convolutional Network for Traffic Flow Prediction | CAFGCN | MAE\RMSE\MAPE | 我博 |
| Spatiotemporal Multi-Graph Convolution Network for Ride-Hailing Demand Forecasting | ST-MGCN | ||
| Bike Flow Prediction with Multi-Graph Convolutional Networks | MGCN | RMSE | 我博 |
| RedPacketBike: A Graph-Based Demand Modeling and Crowd-Driven Station Rebalancing Framework for Bike Sharing Systems | RMSE\MAE | 我博 |