This work considers the task of representation learning on the attributed relational graph (ARG). Both the nodes and edges in an ARG are associated with attributes/features allowing ARGs to encode rich structural information widely observed in real applications. Existing graph neural networks offer limited ability to capture complex interactions within local structural contexts, which hinders them from taking advantage of the expression power of ARGs. We propose Motif Convolution Module (MCM), a new motif-based graph representation learning technique to better utilize local structural information. The ability to handle continuous edge and node features is one of MCM’s advantages over existing motif-based models. MCM builds a motif vocabulary in an unsupervised way and deploys a novel motif convolution operation to extract the local structural context of individual nodes, which is then used to learn higher-level node representations via multilayer perceptron and/or message passing in graph neural networks. When compared with other graph learning approaches to classifying synthetic graphs, our approach is substantially better in capturing structural context. We also demonstrate the performance and explainability advantages of our approach by applying it to several molecular benchmarks.

Particulate matter is a severe source of atmospheric pollution in urban cities, and it has adverse effects on human health. This study was conducted during the whole year of 2016 to monitor the concentrations of PM₁₀ and PM_2.5 on the Beijing Hanshiqiao wetland and bare land in Beijing to analyze their correlations with meteorological factors and compare the removal efficiency between two land surface types. The results indicated that (1) the PM₁₀ and PM_2.5 concentrations on the bare land were higher than those on wetland as a whole, reaching the highest value both at night and dusk and the lowest value near noon. The average concentration of PM₁₀ was higher in winter (wetland: 137.48 μg·m^-3; bare land: 164.75 μg·m^-3) and spring (wetland: 205.18 μg·m^-3; bare land: 244.85 μg·m^-3) and the concentration of PM_2.5 on the wetland also reached the higher value in winter and spring with the average of 84.52 μg·m^-3 and 98.98 μg·m^-3, whereas, it was higher in spring and summer on the bare land; (2) concentrations of PM₁₀ and PM_2.5 were significantly positively affected by the relative humidity (P < 0.01) and negatively influenced by wind speed (P < 0.05). The relationship between PM₁₀ and PM_2.5 concentrations and temperature was found complicated: it showed a significantly negative correlation (P < 0.01) in winter and spring and was insignificant in autumn, but in summer, only the correlation between the PM₁₀ concentration and temperature on wetland was significant (P < 0.01); (3) the removal efficiencies of PM₁₀ and PM_2.5 followed the order of spring > winter > autumn > summer on the wetland, and the removal efficiency of PM₁₀ was greater than that of PM_2.5. This study is aim to provide practical measures to improve the air quality and facilitate sustainable development in Beijing.