Tag
This paper proposes a spatiotemporal graph Transformer framework for traffic forecasting in edge computing, combining graph neural networks for spatial correlations and Transformer self-attention for long-range temporal dependencies. Experiments on real-world cellular data show it outperforms recurrent graph-based baselines like GCN-LSTM and GCN-GRU.
EMAGN is a new efficient multi-attention graph network for traffic forecasting that linearizes self-attention via learned clustering, reducing complexity from quadratic to linear while maintaining accuracy close to full-attention models, with significant reductions in training time, inference time, and GPU memory.
This paper introduces STKAN, a spatio-temporal forecasting architecture that integrates Taylor-polynomial Kolmogorov-Arnold Network modules for spatial and temporal token mixing. Experiments on five traffic benchmarks show competitive performance, suggesting nonlinear function approximators can complement architectural design.
This paper investigates whether complex transformer-based attention is necessary for global spatial information extraction in traffic forecasting, finding that simple global aggregation operators achieve comparable performance with lower computational complexity.
This paper introduces regime-stratified evaluation for time series foundation models, revealing that aggregate metrics hide severe failures during traffic regime transitions, and proposes bimodal mixture augmentation to improve coverage while preserving overall accuracy.
PatchSTG introduces a patch-based spatiotemporal graph Transformer for traffic forecasting on irregular sensor networks, achieving near-linear complexity while maintaining competitive performance.
Proposes GC-MoE, a graph-conditioned mixture of experts framework for traffic forecasting that assigns each node a personalized combination of frozen pretrained spatio-temporal GNN experts based on graph topology and recent input, training only a lightweight routing module (∼17K parameters) and achieving competitive performance on four benchmarks.
Proposes a Global-Local Graph Attention Network (GLGAT) with pairwise encoding and event-based adjacency matrix for traffic forecasting, effectively capturing spatio-temporal correlations and achieving competitive performance on real-world datasets.