Learning spatial-temporal feature with graph product

Earlier works on dynamic spatial-temporal data modelling prefer using spatial-temporal factorized graph convolutional networks (GCNs), which are easy to interpret but fail to capture joint spatial-temporal cor-relations. Thus, lots of subsequent research focus on constructing a localized adjacent matrix to capture joint features from both spatial and temporal dimension simultaneously. However, their ways of building the adjacent matrices are usually heuristic, which makes the models difficult to interpret. Meanwhile, the lack of theoretical explanations hinders the model's generalization. We introduce a general framework to model dynamic spatial-temporal graph data from the view of graph product. With the power of graph product, we propose a systematical way of constructing the spatial-temporal adjacent graphs, which can not only improve the model's interpretability but increase the spatial-temporal receptive field. Under the novel framework, the existing methods can be taken as special cases of our model. Extensive experi-ments on multiple large-scale real-world datasets, NTU-RGB+D60, NTU-RGB+D120, UAV-Human, PEMS03, PEMS04, PEMS07, and PEMS08, demonstrate that the proposed model can generalize to most of the sce-narios with a performance improvement in a significant margin compared to the state-of-the-art meth-ods.(c) 2023 Elsevier B.V. All rights reserved.

Automated summary

Earlier works on dynamic spatial-temporal data modelling used spatial-temporal factorized graph convolutional networks (GCNs), which lack joint spatial-temporal correlations. Subsequent research focused on constructing localized adjacent matrices, but their methods were usually heuristic and lacked interpretability. This study proposed a general framework using graph product to model dynamic spatial-temporal graph data, with a systematic method of constructing spatial-temporal adjacent graphs, improving interpretability and increasing the spatial-temporal receptive field. Experiment results on various real-world datasets demonstrated significant performance improvement compared to state-of-the-art methods.