Vincent Zheng

Be Open-Minded

About Me

Greetings! I am Vincent, a PhD candidate affiliated with McGill University’s Department of Civil Engineering, where I have been actively pursuing my doctoral studies since January 2021. Collaborating closely with Prof. Lijun Sun, my research endeavors revolve around the captivating realms of spatiotemporal modeling and machine learning. In 2017, I obtained my Bachelor’s degree in Transportation Engineering from Central South University, followed by the completion of my Master’s degree in the same field in 2020. Throughout my Master’s research, I delved into the fascinating intricacies of human mobility and complex network analysis, honing my expertise in this captivating domain. I am excited to share with you a curated selection of my academic works and experience here!

A Glimpse of My Work

Here is a collection of my papers that I believe will be worth your time reading. For a comprehensive list of my publications, please feel free to explore my Google Scholar profile. I hope you find these reading suggestions enlightening and enjoy delving into the realms of knowledge they offer.

  • Better Batch for Deep Probabilistic Time Series Forecasting: In this work, I explore error autocorrelation in probabilistic time series forecasting models. Drawing inspiration from the generalized least squares (GLS) method in linear regression, I extend the Gaussian likelihood of a univariate model to a multivariate Gaussian likelihood. By incorporating a time-invariant covariance matrix within a mini-batch, I aim to capture error autocorrelation. This approach enhances the accuracy of estimated distribution parameters during prediction, leveraging the learned covariance matrix to account for past observed errors.
  • Enhancing Deep Traffic Forecasting Models with Dynamic Regression: In this work, I tackle the issue of residual autocorrelation in deterministic time series forecasting using dynamic regression. To capture the autocorrelation, I assume that the residual follows a matrix autoregressive (AR) process. Additionally, I model the concurrent spatiotemporal correlation by employing a matrix normal distribution.
  • Understanding Coupling Dynamics of Public Transportation Networks: The literature review serves as a valuable starting point for those interested in exploring multiplex network theory within the context of transportation. Furthermore, if you are keen on utilizing smart card data for trip reconstruction, the methods section offers insightful guidance. However, the results section, focusing on region-specific analysis, may not be necessary for all readers.


  • I received the FRQNT B2X Scholarship from Fonds de recherche du Québec (FRQ), a package of C$77,000, 2022
  • I received the McGill Engineering Doctoral Award (MEDA), a package of C$111,000, 2021
  • I was accepted by two PhD programs at UW Seattle and McGill, 2020
  • Outstanding Graduate of Hunan Province (Master Student), China, 2020
  • National Scholarship for Graduate Student (China), 2018
  • Outstanding Graduate of Hunan Province (Bachelor Student), China, 2017
  • National Scholarship for Outstanding Students Majoring in Transportation Engineering (China), 2016
  • 1st Place Winner, National Competition of Transport Science and Technology (China), 2016

Teaching Experience

  • TA, CIVE 542 Transportation Network Analysis, Winter 2023, McGill University
  • TA, CIVE 319 Transportation Engineering, Winter 2022, McGill University

Work Experience

ExPretio Technologies

Mitacs Accelerate Intern

July 2021 - Nov. 2021

Project: Adaptive multi-horizon models for probabilistic demand forecasting