我们认为TGConv具有通用性,可以应用于其他任务,并将其留待进一步研究。 3.5 Spatio-Temporal Graph Transformer 时间transformer可以单独模拟每个行人的运动动力学,但不能考虑空间交互作用;spatial Transformer利用TGConv处理人群交互,但很难推广到时间序列。行人预测的一个主要挑战是建模耦合时空交互作用。行人的空间和时间...
Motivation: 首先现有的方法大多针对grid-based和point-based问题,忽略了segment-level的流量预测。其次GCN比较依赖于Laplace矩阵,通常输入图的邻接矩阵是固定的,而实际上道路graph通常具有时变特性,且过去的研究基本都使用地理距离来表达邻接矩阵,实际上地理上的距离并不能很好的体现位置之间的空间相关性。 Preliminaries: ...
To predict long-term dependencies more accurately, in this paper, a new and more effective traffic flow prediction model is proposed.Design/methodology/approachThis paper proposes a new and more effective traffic flow prediction model, named channel attention-based spatial-temporal graph neural networks...
Each space–time block is composed of two graph attention networks and a gated recurrent unit, which are used to extract the spatial and temporal characteristics of road traffic flow respectively, while adding residual connections to prevent the gradient from disappearing. Then, with the traffic ...
Predicting Traffic Flow Using Dynamic Spatial-Temporal Graph Convolution Networks traffic flow.By combining graph attention network and time attention network,a spatiotemporal convolution block is constructed to capture spatiotemporal ... Y Liu,F Wan,C Liang - 《Computers Materials & Continua》 被引量...
ImputeFormer: Low Rankness-Induced Transformers for Generalizable Spatiotemporal Imputation Pre-Training Identification of Graph Winning Tickets in Adaptive Spatial-Temporal Graph Neural Networks 其它 DyPS: Dynamic Parameter Sharing in Multi-Agent Reinforcement Learning for Spatio-Temporal Resource Allocation ...
In light of the positive results presented in this work, as well as its present limitations, we can foresee several research avenues to improve spatio-temporal models further. First, using time-varying, as opposed to static networks, either through auto-regressive or recurrent architectures, would...
In this paper, we propose a novel deep learning framework, Spatio-Temporal Graph Convolutional Networks (STGCN), to tackle the time series prediction problem in traffic domain. Instead of applying regular convolutional and recurrent units, we formulate the problem on graphs and build the...
图一、STAR中的temporal transformer以及spatial transformer。第一眼看上去右边这个和GAT很像。 2、模型设计 构建图:空间图中的边表示两个行人之间的距离小于一定阈值。 针对Temporal Transformer,就正常用Transformer即可。 【KEY】针对Spatial Transformer 首先有一个观察:self-attention可以看作是在无向全连图上传递信...
ST-Meta Graph Reconstruction进一步设计用于通过重建不同城市的结构关系来进行结构感知元训练。 ST-GFSL 的端到端学习过程遵循基于MAML的episode learning。通过模拟目标城市的小样本场景,对批量的小样本训练任务进行采样,得到适应性强的基础模型。 Spatio-Temporal Neural Network ...