Graph representation learning for road type classification

作者：

Highlights：

• Learning to classify road networks of realistic cities from open street map.

• Applying line graph transformation to use qualitative road segment fea- tures.

• Proposing a sampling function of nodes in local and global topological neighborhoods.

• Proposing a novel approach to aggregation, Graph Attention Isomorphism Network, GAIN.

摘要

•Learning to classify road networks of realistic cities from open street map.•Applying line graph transformation to use qualitative road segment fea- tures.•Proposing a sampling function of nodes in local and global topological neighborhoods.•Proposing a novel approach to aggregation, Graph Attention Isomorphism Network, GAIN.

论文关键词：Road network graphs,Graph representation learning,Line graph transformation,Neighborhood aggregation,Topological neighborhood

论文评审过程：Received 17 February 2021, Revised 21 May 2021, Accepted 6 July 2021, Available online 15 July 2021, Version of Record 25 July 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108174