Introduction
Road networks are naturally dynamic out of both the predictable factors (like traffic flow during commuting time, seasonal tourism and commercial activities) and unpredictable factors (like sudden weather change, traffic accidents). How to effectively represent and cluster them is beneficial to multiple downstream tasks including but not limited to route optimization, traffic prediction, urban planning and location-based services. Even though various graph embedding methods have been proposed, most of them are node-level or edge-level embedding, which cannot be used to solve our question. In addition, many are designed for the topological or label-based embedding. Nevertheless, in our setting, the road network for one city has few changes on the topology or associated with the labels and most changes happen on the travelling time along road segments (i.e. edge weight). Moreover, existing graph embedding methods can hardly handle large-scale networks (for instance with hundreds of thousands of vertices). In this project, we focus on the graph-level embedding of road network with challenges in dealing with the edge weight changes and scalability.
Research tasks
We will implement this project with the following subtasks:
- Familiarize with the existing graph-level network embedding methods
- Collect the road network dataset
- Design graph embedding model with good representation
- Improve the scalability of embedding model and apply it in large road networks
- Test the proposed method and compare it with the state-of-the-art
Reference
[1] Narayanan, Annamalai, Mahinthan Chandramohan, Rajasekar Venkatesan, Lihui Chen, Yang Liu, and Shantanu Jaiswal. “graph2vec: Learning distributed representations of graphs.” arXiv preprint arXiv:1707.05005 (2017).
[2] You, Yuning, Tianlong Chen, Yang Shen, and Zhangyang Wang. “Graph contrastive learning automated.” In International Conference on Machine Learning, pp. 12121-12132. PMLR, 2021.
[3] Chang, Yanchuan, Egemen Tanin, Xin Cao, and Jianzhong Qi. “Spatial Structure-Aware Road Network Embedding via Graph Contrastive Learning.” In EDBT, pp. 144-156. 2023.
Requirement
Background and experience in graph embedding (preferred), ML and graph theory.
Programming experience in Python is essential.
Contact
If you are interested in this project, contact Dr. Mengxuan Zhang