News

2024

November 20: It is a great pleasure to have Dr. Mehrdad Ghadiri from MIT Sloan School of Management as our guest speaker, sharing his work on "Approximately Optimal Core Shapes for Tensor Decompositions" at the SRB seminar of MIT Urban Mobility Lab.
November 13: It is a great pleasure to have Dr. T. Konstantin Rusch from MIT CSAIL Lab as our guest speaker, sharing his recent work on "Message-Passing Monte Carlo: Generating low-discrepancy point sets via graph neural networks" (the latest paper published in PNAS) at the SRB seminar of MIT Urban Mobility Lab.
November 13: It is a great pleasure to have Prof. Filipe Rodrigues from the Machine Learning for Smart Mobility group at the Technical University of Denmark (DTU) as our guest speaker, sharing his recent work on "Reinforcement Learning for Network Optimization in Transportation" at the SRB seminar of MIT Urban Mobility Lab.
November 6: It is a great pleasure to have Prof. Qiusheng Wu from the Department of Geography & Sustainability at the University of Tennessee, Knoxville (UTK) as our guest speaker, sharing his substantial work on "Open-source Python package and software development in geospatial data science" at the SRB seminar of MIT Urban Mobility Lab.
October 18: I will host the invited session at 2024 INFORMS Annual Meeting on October 21th in Seattle, USA. The topic is "Simulation and learning for smart transportation systems". In this session, I will give a talk about "Modeling urban traffic data with matrix and tensor approaches" (check out the slides).
October 16: It is a great pleasure to have Prof. Vassilis Digalakis Jr from the Information Systems and Operations Management Department at HEC Paris as our guest speaker, sharing his recent work on "Structural stability in machine learning" (including the latest paper "Slowly varying regression under sparsity" published in Operations Research) at the SRB seminar of MIT Urban Mobility Lab.
September 2: I revisit three very common methods for data dimensionality reduction by watching Latent Space Visualisation: PCA, t-SNE, UMAP | Deep Learning Animated on YouTube. I really like this kind of explanation approach.
August 19: Our paper "Forecasting urban traffic states with sparse data using Hankel temporal matrix factorization" (authors: Xinyu Chen, Xi-Le Zhao, Chun Cheng) was accepted to INFORMS Journal on Computing. This paper received a lot of insightful feedback and suggestions from reviewers. Thank you for the great contributions to this work in the reviewing process.
June 23: Our paper "Laplacian convolutional representation for traffic time series imputation" (authors: Xinyu Chen, Zhanhong Cheng, HanQin Cai, Nicolas Saunier, Lijun Sun) was accepted to IEEE Transactions on Knowledge and Data Engineering. This paper received a lot of insightful feedback and suggestions from reviewers. Thank you for the great contributions to this work in the reviewing process.
May 16: I am delighted to give a talk at Northeastern University (invited by Prof. Ryan Qi Wang with the Department of Civil and Environmental Engineering) in Boston, USA. The topic is about "Modeling temporal correlations and dynamics in spatiotemporal data systems", highlighting the importance of temporal modeling (e.g., local time series trends and time-varying system behaviors) in real-world data.
March 31: We provide slides for understanding the most basic definition, properties, and derivatives of matrix traces in matrix computations and machine learning.
March 20: I gave a comment (i.e., Rethink equation (15) for the iterative shrinkage thresholding algorithm) for Attention: Self-expression is all you need on OpenReview. This work has tried to connect transformers with the prior art, a very good attempt!
March 15: I moved from Montreal to Boston, starting a new position as Postdoctoral Associate at MIT with Prof. Jinhua Zhao.
February 13: I changed my homepage photo as this one, which was took at Casa Loma, Toronto, Canada.
February 12: Our papers have received 1,000 citations on Google Scholar. Many thanks to the research community!
January 11: I gave a presentation with my supervisor Prof. Nicolas Saunier about "Open-source project: Machine learning for transportation data imputation and prediction" (check out the slides) at the Reproducible Research Workshop of the 103rd Transportation Research Board (TRB) Annual Meeting in Washington D.C., USA.

2023

December 28: I gave a virtual talk about "Matrix and tensor methods for spatiotemporal traffic data imputation and forecasting" (check out the slides) at Southern University of Science and Technology (SUSTech) Forum with the Department of Statistics and Data Science.
December 13: We write a blog post for understanding and reproducing our paper "Discovering dynamic patterns from spatiotemporal data with time-varying low-rank autoregression". The content includes an intuitive explanation of time-varying autoregression, spatiotemporal data processing, and the algorithm implementation with NumPy/CuPy. It would be a good material for beginners to follow our research. Any feedback would be highly appreciated.
December 11: I successfully passed my PhD research defense today. It would be my great pleasure to have Prof. Francesco Ciari, Prof. Nicolas Saunier, Prof. Lijun Sun, Prof. James Goulet, and Prof. Guillaume Rabusseau as the jury members. If you are interested in my PhD research, welcome to take a look at the slides. Special thanks to my supervisor Prof. Nicolas Saunier for the great suggestion when I was organizing all materials.
October 5: I revisit the core concept of linear algebra by watching Essence of Linear Algebra on YouTube. I really like the visuals-first approach for explaining linear algebra in this series.
September 11: Four of our papers have been cited above 100 times on Google Scholar, including all three (first-author) papers published in Transportation Research Part C: Emerging Technologies during my Master's study.
July 25: Our papers have received 800 citations on Google Scholar. Many thanks to the research community!
July 17: I will give a talk about "Laplacian convolutional representation for traffic time series imputation" (check out the slides) at the WCTR conference on July 19th in Montreal, Canada.
July 4: Our open-source project transdim got 1,000+ stars and 270+ forks on GitHub. This project focuses on providing machine learning solutions for transportation data imputation and prediction. It is really a long journey to reach here since the first commit on September 23, 2018.
July 2: Our paper "Discovering dynamic patterns from spatiotemporal data with time-varying low-rank autoregression" (authors: Xinyu Chen, Chengyuan Zhang, Xiaoxu Chen, Nicolas Saunier, Lijun Sun) was accepted to IEEE Transactions on Knowledge and Data Engineering. Chengyuan Zhang was contributed equally to this work, thank you!
May 22: I am delighted to give a talk at Southern University of Science and Technology (南方科技大学, invited by Prof. Lili Yang with Department of Statistics and Data Science) in Shenzhen with the topic "Low-rank matrix and tensor methods for spatiotemporal traffic data modeling" (check out the slides).
May 16: Our papers have received 700 citations on Google Scholar. Many thanks to the research community!
April 21: I am delighted to give a talk at University of Electronic Science and Technology of China (电子科技大学, invited by Prof. Xi-Le Zhao with School of Mathematical Science) in Chengdu with the topic "Low-rank matrix and tensor methods for spatiotemporal data modeling" (check out the slides).
April 20: I am delighted to give a talk at Sichuan University (四川大学, invited by Prof. Yuankai Wu) in Chengdu with the topic "Low-rank matrix and tensor methods for spatiotemporal data modeling" (check out the slides).
March 15: I am delighted to give a (virtual) talk at Soochow University with the topic "Low-rank matrix and tensor factorization for speed field reconstruction" (check out the slides or the latest version).
March 10: Our paper "Bayesian temporal factorization for multidimensional time series prediction" (2022) published in IEEE Transactions on Pattern Analysis and Machine Intelligence was selected as a hot paper (in the top 0.1% of papers in the academic field of Engineering (proof)) from Essential Science Indicators (ESI) Data.
March 9: I gave a presentation about "Low-rank matrix and tensor factorization for speed field reconstruction" at the Research Group of Transport, Polytechnique Montreal. If you are interested, welcome to take a look at the slides and the blog post!
March 8: Our open-source project awesome-latex-drawing got 1,000+ stars and 130+ forks on GitHub. This project is a collection of 30+ academic drawing examples for using LaTeX, including Bayesian networks, function plotting, graphical models, tensor structure, and technical frameworks.
February 22: Our GitHub repositories accumulated more than 3,000 stars.
January 26: Our papers have received 600 citations on Google Scholar. Many thanks to the research community!
January 22: Our open-source project transdim got 900+ stars and 260+ forks on GitHub. This project focuses on providing machine learning solutions for transportation data imputation and prediction.
January 21: We posted an intuitive understanding of tensors in machine learning on Medium. The content includes the basic components (entry/fiber/slice) of third-order tensors and some tensor examples in real-world applications. If you have any suggestions or feedback, feel free to contact us.
January 13: Our paper "Bayesian temporal factorization for multidimensional time series prediction" (2022) published in IEEE Transactions on Pattern Analysis and Machine Intelligence was selected as a highly cited paper (in the top 1% of papers in the academic field of Engineering (proof)) from Essential Science Indicators (ESI) Data.
January 11: I gave a poster presentation about "Nonstationary temporal matrix factorization for sparse traffic time series forecasting" at TRB 2023 in Washington, D.C., USA.

2022

December 5: The preprint of our recent work "Lapalacian convolutional representation for traffic time series imputation" (authors: Xinyu Chen, Zhanhong Cheng, Nicolas Saunier, Lijun Sun) is available at arXiv. The adapted datasets and Python implementation are available at https://github.com/xinychen/transdim. For reproducing the experiment of univariate time series imputation, please check out the folder transdim/tree/master/univariate-models. For reproducing the experiment of multivariate time series imputation, please check out the folder transdim/tree/master/multiviarate-models.
November 28: The preprint of our recent work "Discovering dynamic patterns from spatiotemporal data with time-varying low-rank autoregression" (authors: Xinyu Chen, Chengyuan Zhang, Xiaoxu Chen, Nicolas Saunier, Lijun Sun) is available at arXiv. The adapted datasets and Python implementation are available at https://github.com/xinychen/vars.
October 19: Our blog post "Temporal matrix factorization for multivariate time series forecasting" has reached 4,000 views on Medium.
October 10: Our papers have received 500 citations on Google Scholar. Many thanks to the research community!
September 26: Our open-source project latex-cookbook got 500+ stars and 80+ forks on GitHub. This project is about "Academic writing with LaTeX: A tutorial (in Chinese)" (《LaTeX论文写作教程》).
September 1: Our paper "Bayesian temporal factorization for multidimensional time series prediction" (2022) published in IEEE Transactions on Pattern Analysis and Machine Intelligence was included among the 50 most frequently accessed documents for the journal. Take a look at the popular article feed (August 2022).
August 5: Our paper "Low-rank autoregressive tensor completion for spatiotemporal traffic data imputation" (2022) published in IEEE Transactions on Intelligent Transportation Systems was selected as a hot paper (in the top 0.1% of papers in the academic field of Engineering (proof)) from Essential Science Indicators (ESI) Data.
August 5: Our paper "A Bayesian tensor decomposition approach for spatiotemporal traffic data imputation" (2019) published in Transportation Research Part C: Emerging Technologies was selected as a highly cited paper (in the top 1% of papers in the academic field of Engineering (proof)) from Essential Science Indicators (ESI) Data.
July 22: Our open-source project transdim got 800+ stars and 240+ forks on GitHub. This project focuses on providing machine learning solutions for transportation data imputation and prediction.
July 22: Our GitHub repositories accumulated more than 2,500 stars.
June 20: Our blog post "Implementing Kronecker product decomposition with NumPy" is available on Medium.
June 18: Our blog post "Visualizing global sea surface temperature data in Python" is available on Medium.
May 24: I gave a summary of my Ph.D. research project about spatiotemporal traffic data imputation and forecasting with tensor learning at the IVADO Project Workshop. If you are interested, welcome to take a look at the slides and give me feedback!
May 18: Our papers have received 400 citations on Google Scholar. Many thanks to the research community!
May 11: I am delighted to give a (virtual) talk at Hong Kong Machine Learning Meetup with the topic "Nonstationary temporal matrix factorization for multivariate time series forecasting". The slides are available at https://xinychen.github.io/slides/notmf.pdf.
April 25: Our blog post "Forecasting multivariate time series with nonstationary temporal matrix factorization" is available on Medium.
April 14: Our open-source project awesome-latex-drawing got 800+ stars and 120+ forks on GitHub. This project is a collection of 20+ academic drawing examples for using LaTeX, including Bayesian networks, function plotting, graphical models, tensor structure, and technical frameworks.
April 7: I am delighted to give a (virtual) talk at Guangxi University with the topic "Nonstationary temporal matrix factorization for multivariate time series forecasting". The slides are available at https://xinychen.github.io/slides/notmf.pdf.
March 21: The preprint of our recent work "Nonstationary temporal matrix factorization for multivariate time series forecasting" (authors: Xinyu Chen, Chengyuan Zhang, Xi-Le Zhao, Nicolas Saunier, Lijun Sun) is available at arXiv. The adapted datasets and Python implementation (both CPU and GPU) are available at https://github.com/xinychen/tracebase. For an intuitive explanation, we provide a blog post "Temporal matrix factorization for multivariate time series forecasting" on Medium.
March 15: Our blog post "Inpainting fluid dynamics with tensor decomposition (NumPy)" is available on Medium. The Python codes in this blog post are available at https://github.com/xinychen/fluid-inpainting.
February 23: Our blog post "Using conjugate gradient to solve matrix equations" is available on Medium. The source LaTeX/Python codes in this blog post are available at https://github.com/xinychen/conjugate-gradient.
February 17: Our blog post "Intuitive understanding of Newton-Raphson method" is available on Medium. The source LaTeX codes for creating graphics are available at https://github.com/xinychen/Newton-Raphson.
February 14: Our blog post "Analyzing missing data problem in Uber movement speed data" is available on Medium. The Python codes in this blog post are available at https://github.com/xinychen/tracebase.
January 26: Our GitHub repositories accumulated more than 2,000 stars.
January 11: Our papers have received 300 citations on Google Scholar. That is a new milestone for my research career, many thanks to the research community!
January 9: Our open-source project latex-cookbook got 200+ stars and 30+ forks on GitHub. This project is about "Academic writing with LaTeX: A tutorial (in Chinese)" (《LaTeX论文写作教程》).

2021

December 16: I am delighted to be awarded the CIRRELT PhD Excellence Scholarship. For more details on Interuniversity Research Centre on Enterprise Networks, Logistics and Transportation (CIRRELT) and its mission, please check out the website.
December 2: Our open-source project transdim got 700+ stars and 220+ forks on GitHub.
October: Our blog post "Dynamic mode decomposition for spatiotemporal traffic speed time series in Seattle freeway" was available on Medium.
October: Our GitHub repository awesome-latex-drawing got 700 stars. In this open-source project, we provided several LaTeX drawing examples that are from our publications.
October: Our blog post "Reduced-rank vector autoregressive model for high-dimensional time series forecasting" was available on Medium.
October: Our blog post "Dynamic mode decomposition for multivariate time series forecasting" was available on Medium.
October: Our blog post "Matrix autoregressive model for multidimensional time series forecasting" was available on Medium.
July: Our paper "Low-rank autoregressive tensor completion for spatiotemporal traffic data imputation" (authors: Xinyu Chen, Mengying Lei, Nicolas Saunier, Lijun Sun) was accepted to IEEE Transactions on Intelligent Transportation Systems.
June: Our open-source project transdim got 600+ stars and 190+ forks on GitHub.
May: Our paper "Scalable low-rank tensor learning for spatiotemporal traffic data imputation" (authors: Xinyu Chen, Yixian Chen, Nicolas Saunier, Lijun Sun) was accepted to Transportation Research Part C: Emerging Technologies.
May: Our GitHub repositories accumulated more than 1,500 stars.
March: Our paper "Bayesian temporal factorization for multidimensional time series prediction" (authors: Xinyu Chen, Lijun Sun) was accepted to IEEE Transactions on Pattern Analysis and Machine Intelligence.
February: Our open-source project transdim got 500+ stars and 150+ forks on GitHub.

2020

July: Our GitHub repositories accumulated more than 1,000 stars.
July: Our blog post "Intuitive understanding of randomized singular value decomposition: A Python implementation of SVD with randomized linear algebra" was available on Medium.
June: Our paper "Low-rank autoregressive tensor completion for multivariate time series forecasting" was available on arXiv.
April: I am delighted to be awarded the IVADO PhD Excellence Scholarship. The awarded research topic is "City-Scale Traffic Data Imputation and Forecasting with Tensor Learning". For more details on Institute for Data Valorisation (IVADO) and its mission, please check out the website.