Download Solar-Energy, Traffic, Electricity, Exchange-rate datasets fromhttps://github.com/laiguokun/multivariate-time-series-data. Uncompress them and move them to the data folder. Download the METR-LA and PEMS-BAY dataset fromGoogle DriveorBaidu Yunprovided byLi et al.. Move them into the ...
图网络 MTGNN 简介1 文章概要 论文:Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks 下载:https://arxiv.org/abs/2005.11650 代码:https://github.com/nnzhan/MTGNN 2 …
源码链接 |github.com/nnzhan/MTGNN Overview 本此分享的论文是时空图神经网络系列的第一篇论文,来自KDD2020。本文将对论文中提出的几个核心模块代码的个人解读进行分享。 多元时间序列预测背后的一个基本假设是其变量相互依赖,然而现有的方法未能充分利用变量之间潜在的空间依赖关系。图神经网络(GNN)在处理关系依赖方面...
源码链接 |https://github.com/nnzhan/MTGNN ❞ TL;DR 考虑到多变量间存在的依赖关系,论文中针对多变量时序预测问题提出了一个通用的图神经网络框架 MTGNN。首先可以通过graph learning模块学习到变量间的单向关系构图,然后通过graph convolution模块和temporal convolution模块来捕获时间序列间的时空依赖特征,此外这三个...
Time series forecasting especially in LSTF compare,include Informer, Autoformer, Reformer, Pyraformer, FEDformer, Transformer, MTGNN, LSTNet, Graph WaveNet - Masterleia/TSF_LSTF_Compare
Download Solar-Energy, Traffic, Electricity, Exchange-rate datasets fromhttps://github.com/laiguokun/multivariate-time-series-data. Uncompress them and move them to the data folder. Traffic datasets Download the METR-LA and PEMS-BAY dataset fromGoogle DriveorBaidu Yunprovided byLi et al.. Move...