Hongjun Wang

Ph.D. Candidate, The University of Tokyo

I'm currently a Ph.D. Candidate at Department of Mechano-Informatics, The University of Tokyo, supervised by Prof. Yinqiang Zheng. Prior to this, I attained my master's degree, supervised by Prof. Xuan Song from the Department of Computer Science at Southern University of Science and Technology. I earned my B.S. degree in Information and Computer Science from Nanjing University of Posts and Telecommunications.

My research interests focus on spatiotemporal modeling, traffic prediction, and urban computing. I work on developing efficient graph neural networks and transformer architectures for understanding and predicting complex urban dynamics.


Research Summary. My research focuses on developing advanced machine learning models for spatiotemporal data analysis, particularly in urban computing and transportation systems. I have contributed to several key areas:
  1. Spatiotemporal Graph Neural Networks and Urban Computing. I developed STGformer, an efficient spatiotemporal graph transformer for traffic forecasting, and work on evaluating generalization ability across different urban environments. I also contributed to GOF-TTE, a generative online federated learning framework for travel time estimation, and work on multi-task weakly supervised learning approaches. [IEEE TMC 2025] [IEEE TKDE 2022] [AAAI 2023]
  2. Computer Vision and Causal Analysis. I developed novel approaches for image super-resolution and generalizable computer vision models, and work on causal-based supervision of attention mechanisms in graph neural networks. [CVPR 2024] [ICCV 2025] [IJCAI 2023] [IEEE T-ITS 2023]
  3. Environmental AI and Visual Analytics. I work on AI applications for environmental sciences, particularly flood forecasting and natural hazard prediction, and develop visual analytics approaches for interpreting traffic prediction models. [npj Natural Hazards 2025] [CIKM 2023] [CVMJ 2023] [Sensors 2023]

Internships

Publications

2025

Not All Degradations Are Equal: A Targeted Feature Denoising Framework for Generalizable Image Super-Resolution

Hongjun Wang, Jiyuan Chen, Zhengwei Yin, Xuan Song, Yinqiang Zheng

ICCV 2025 - paper

We presents a general solution that can be seamlessly integrated with existing super-resolution models without requiring architectural modifications.

Evaluating the Generalization Ability of Spatiotemporal Model in Urban Scenario

Hongjun Wang, Jiyuan Chen, Tong Pan, Zheng Dong, Lingyu Zhang, Renhe Jiang, Xuan Song

IEEE Transactions on Mobile Computing (TMC) 2025 - paper | code

A comprehensive study on the generalization challenges of spatiotemporal models in urban computing scenarios.

2024

Navigating Beyond Dropout: An Intriguing Solution Towards Generalizable Image Super Resolution

Hongjun Wang, Jiyuan Chen, Yinqiang Zheng, Tieyong Zeng

CVPR 2024 - paper | code

We propose a novel approach for generalizable image super-resolution that goes beyond traditional dropout techniques.

HoLens: A visual analytics design for higher-order movement modeling and visualization

Zezheng Feng, Fang Zhu, Hongjun Wang, Jianing Hao, Shuang-Hua Yang, Wei Zeng, Huamin Qu

Computational Visual Media (CVMJ) 2024 - paper

A visual analytics framework for modeling and visualizing higher-order movement patterns in urban environments.

2023

Assessing the Continuous Causal Responses of Typhoon-related Weather on Human Mobility: An Empirical Study in Japan

Zhang Zhiwen, Hongjun Wang, Zipei Fan, Ryosuke Shibasaki, Xuan Song

CIKM 2023 - paper

An empirical study analyzing the causal effects of typhoon-related weather on human mobility patterns in Japan.

ST-ExpertNet: A Deep Expert Framework for Traffic Prediction

Hongjun Wang, Jiyuan Chen, Zipei Fan, Zekun Cai, Ryosuke Shibasaki, Xuan Song

IEEE Transactions on Knowledge and Data Engineering (TKDE) 2023 - paper

A deep expert framework that leverages multiple specialized models for improved traffic prediction accuracy.

TrafPS: A shapley-based visual analytics approach to interpret traffic

Zezheng Feng, Yifan Jiang, Hongjun Wang, Zipei Fan, Shuang-Hua Yang, Huamin Qu, and Xuan Song

Computational Visual Media (CVMJ) 2023 - paper

A Shapley-based visual analytics approach for interpreting traffic prediction models and understanding feature contributions.

Causal-Based Supervision of Attention in Graph Neural Network: A Better and Simpler Choice towards Powerful Attention

Hongjun Wang, Jiyuan Chen, Lun Du, Qiang Fu, Xuan Song

IJCAI 2023 - paper

We propose a causal-based supervision approach for attention mechanisms in graph neural networks that leads to more powerful and interpretable attention.

Missing Road Condition Imputation Using a Multi view Heterogeneous Graph Network from GPS Trajectory

Zhiwen Zhang, Hongjun Wang, Zipei Fan, Xuan Song, and Ryosuke Shibasaki

IEEE Transactions on Intelligent Transportation Systems (T-ITS) 2023 - paper

A multi-view heterogeneous graph network approach for imputing missing road condition information from GPS trajectory data.

Easy Begun is Half Done: Spatial-Temporal Graph Modeling with ST-Curriculum Dropout

Hongjun Wang, Jiyuan Chen, Tong Pan, Zipei Fan, Boyuan Zhang, Renhe Jiang, Xuan Song

AAAI 2023 - paper

We introduce ST-Curriculum Dropout, a curriculum learning approach for spatial-temporal graph modeling that improves training efficiency and model performance.

2022

GOF-TTE: Generative Online Federated Learning Framework for Travel Time Estimation

Zhiwen Zhang, Hongjun Wang, Zipei Fan, Jiyuan Chen, Xuan Song, Ryosuke Shibasaki

IEEE Internet of Things Journal (IOT-J) 2022 - paper

A generative online federated learning framework for travel time estimation that preserves privacy while improving prediction accuracy.

Discovering Key Sub-trajectories To Explain the Traffic Prediction

Hongjun Wang, Zipei Fan, Jiyuan Chen, Lingyu Zhang, Xuan Song

Sensors 2022 - paper

A method for discovering key sub-trajectories that are most influential for traffic prediction, providing interpretability for spatiotemporal models.

Multi‑task Weakly Supervised Learning for Origin–Destination Travel Time Estimation

Hongjun Wang, Zhiwen Zhang, Zipei Fan, Jiyuan Chen, Xuan Song, Ryosuke Shibasaki

IEEE Transactions on Knowledge and Data Engineering (TKDE) 2022 - paper

A multi-task weakly supervised learning approach for origin-destination travel time estimation that leverages multiple related tasks.

Multi-modal graph interaction for multi-graph convolution network in urban spatiotemporal forecasting

Lingyu Zhang, Xu Geng, Zhiwei Qin, Hongjun Wang, Xuan Song, Yunhai Wang

Sustainability 2022 - paper

A multi-modal graph interaction approach for multi-graph convolution networks in urban spatiotemporal forecasting.

Awards & Scholarships

  • SPRING GX | The University of Tokyo | Oct 2023 - Present
  • SUSTech Fellowship Program | Southern University of Science and Technology | Oct 2023 - Present
  • Shibasaki Scholarship | The University of Tokyo | Oct 2024 - Present
  • Star of Tomorrow Internship Program | Microsoft Research Asia | 2022-2023