Bio

Dr. Xinyu Chen (陈新宇) is now a Postdoctoral Associate at MIT, working with Prof. Jinhua Zhao on addressing a series of data-driven machine learning problems. He is currently involved in the Mens, Manus, and Machina (M3S) project and the Department of Energy (DOE) projects. Prior to joining MIT, he received his Ph.D. degree from the University of Montreal in Canada where he was fortunately supported by the IVADO PhD Excellence Scholarship from the Montreal Institute for Data Valorization. His Ph.D. thesis was entitled “Matrix and tensor models for spatiotemporal traffic data imputation and forecasting”. Currently, he focuses on developing theoretical and interpretable machine learning algorithms (e.g., unsupervised learning) for modeling a wide range of spatiotemporal data and (computational) social science data. These data are by nature multidimensional tensors collected from real-world systems, including human mobility, trajectory data, traffic flow, fluid flow, climate/weather data, energy consumption data, and international trade data. His research works have been published in the top-tier scientific journals of the fields of Data Science, Machine Learning, Optimization, and Transportation, including

• Xinyu Chen, Xi-Le Zhao, Chun Cheng (2024). Forecasting urban traffic states with sparse data using Hankel temporal matrix factorization. INFORMS Journal on Computing. [PDF] [Data & Python code]
• Xinyu Chen, Zhanhong Cheng, HanQin Cai, Nicolas Saunier, Lijun Sun (2024). Laplacian convolutional representation for traffic time series imputation. IEEE Transactions on Knowledge and Data Engineering, 36 (11): 6490 – 6502. [PDF] [Slides] [Blog post] [Data & Python code]
• Xinyu Chen, Chengyuan Zhang, Xiaoxu Chen, Nicolas Saunier, Lijun Sun (2024). Discovering dynamic patterns from spatiotemporal data with time-varying low-rank autoregression. IEEE Transactions on Knowledge and Data Engineering, 36 (2): 504-517. [PDF] [Blog post] [Data & Python code]
• Xinyu Chen, Lijun Sun (2022). Bayesian temporal factorization for multidimensional time series prediction. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44 (9): 4659-4673. [Slides] [Data & Python code]

All his research papers (~15) have been cited more than 1,300 times on Google Scholar. He is now leading the Spatiotemporal Data Modeling project and addressing many scientific, mathematical, industrial, and engineering problems across the areas of data science, machine learning, and AI for science. He is also a strong advocate of reproducible research and an active developer on GitHub with several popular repositories, such as transdim (1.2k stars) and awesome-latex-drawing (1.3k stars).