Renjie Liao   I am an Assistant Professor (since Jan. 2022) in ECE and CS (associate member) Departments at UBC. I am a Faculty Member at Vector Institute and a Canada CIFAR AI Chair. I was a Visiting Faculty Researcher at Google Brain, working with Geoffrey Hinton and David Fleet. I received my PhD from UofT in 2021, advised by Richard Zemel and Raquel Urtasun. During my PhD, I also worked as a Senior Research Scientist at Uber ATG. I obtained my MPhil from CUHK, advised by Jiaya Jia, and my BEng from Beihang. Since the number of updates to this website each year follows a Poisson distribution (λ ≈ 0.5), please check my Deep Structured Learning Lab website for up-to-date information!

CV

How to Join

Email: renjie.liao@ubc.ca

"There is nothing as practical as a good theory." - Kurt Lewin "The test of all knowledge is experiment." - Richard Feynman

Research Summary:

My current research primarily focuses on probabilistic and geometric deep learning.
In general, I enjoy solving interesting (often practical) problems with principled mathematics and simple implementations.
Below are some representative topics and publications.
• Deep Generative Models: Discrete Diffusion on Symmetric Groups (ICLR 2025), Neural Merton Jump Diffusion (NeurIPS 2025), Discrete Flow Matching for Retrosynthesis (NeurIPS 2025), Gaussian-Bernoulli RBMs, GRAN (NeurIPS 2019)
• Geometric Deep Learning: Latent Graph VAE (ICML 2024), Specformer (ICLR 2023), LanczosNet (ICLR 2019), 3DGNN (ICCV 2017)
• Neural Algorithmic/Math Reasoning: AoPS Math Benchmark (ICML 2025), OOD Generalization in Algorithmic Reasoning (TMLR 2022)
• Generalization Bounds: Information-theoretic OOD Generalization (IEEE TIT 2025, ISIT 2024), PAC-Bayes Bounds for GNNs (ICLR 2021)
• Optimization: Implicit differentiation (ICML 2018), Differentiable Bipartite Matching (ICCV 2019)
• Computer Vision:
  Learning Latent Part-Whole Hierarchy for 3D Point Clouds: CVPR 2024, IJCV 2025
  Image/Video Segmentation: UPSNet (CVPR 2019), DMMNet (ICCV 2019)
  Video Super-resolution: ICCV 2017, ICCV 2015, CVPR 2015
• Self-driving: TrajFlow (IROS 2025), MoFlow (CVPR 2025), LookOut (ICCV 2021), LaneRCNN (IROS 2021), LaneGCN (ECCV 2020), ILVM (ECCV 2020), DSDNet (ECCV 2020), SpAGNN (ICRA 2020)
• Healthcare: GNNs for Left Ventricle Landmark Detection in Echocardiograms (MICCAI 2023), GNNs for Ejection Fraction Estimation in Echocardiograms (MICCAI 2022)

Last Updated: Oct. 15, 2025