Chence Shi 「史晨策」

I am a final-year PhD student at Montreal Institute for Learning Algorithms (Mila) supervised by Prof. Jian Tang. Before that, I received my B.S. in Computer Science from Peking University, advised by Prof. Ming Zhang.

Email: chence.shi [at] umontreal [dot] ca

Github   /   Google Scholar   /   Portfolio     

profile photo
Research Interests

My research is centered on the intersection of generative models, geometric deep learning, graph representation learning, and AI for science.
I am particularly focused on modeling highly multi-modal structured data, such as graphs and proteins.
I aim to explore the tokenize and atomize everything approach to model complex interactions and generative processes within atomistic systems, paving the way for new therapeutic discoveries.

Selected Publications
Protein Sequence and Structure Co-Design with Equivariant Translation
Chence Shi,  Chuanrui Wang,  Jiarui Lu,  Bozitao Zhong,  Jian Tang 
11th International Conference on Learning Representations (ICLR 2023)  
Sequence-structure Co-Design; translation in joint sequence-structure space.
[PDF]    [Code]
E3Bind: An End-to-End Equivariant Network for Protein-Ligand Docking
Yangtian Zhang*,  Huiyu Cai*,  Chence Shi,  Bozitao Zhong,  Jian Tang 
11th International Conference on Learning Representations (ICLR 2023)  
End-to-end protein-ligand docking; SE(3)-equivariance.
[PDF]    [Code]
Learning Gradient Fields for Molecular Conformation Generation
Chence Shi*,  Shitong Luo*,  Minkai Xu,  Jian Tang 
38th International Conference on Machine Learning (ICML 2021)  
Molecular 3D geometry generation; denoising diffusion; SE(3)-equivariance.
Long talk [top 3.0%]
[PDF]    [Code]    [Slides]
TorchDrug: A Powerful and Flexible Machine Learning Platform for Drug Discovery
Zhaocheng Zhu,  Chence Shi,  Zuobai Zhang,  Shengchao Liu,  Minghao Xu,  Xinyu Yuan,  Yangtian Zhang,  Junkun Chen,  Huiyu Cai,  Jiarui Lu,  Chang Ma,  Runcheng Liu,  Louis-Pascal Xhonneux,  Meng Qu,  Jian Tang 
Ready-to-use platfrom for fast development of drug discovery models; ML system.
Deep Learning Library included in PyTorch’s ecosystem
[PDF]     [Homepage]    [Github]    [Twitter]    [Google Colab]
GeoDiff: A Geometric Diffusion Model for Molecular Conformation Generation
Minkai Xu,  Lantao Yu,  Yang Song,  Chence Shi,  Stefano Ermon,  Jian Tang 
10th International Conference on Learning Representations (ICLR 2022)  
Geometric probabilistic models; Markov chains; SE(3)-equivariance; denoising diffusion.
Oral Presentation [54/3391]
[PDF]    [Code]
Predicting Molecular Conformation via Dynamic Graph Score Matching
Shitong Luo*,  Chence Shi*,  Minkai Xu,  Jian Tang 
35th Conference on Neural Information Processing Systems (NeurIPS 2021)  
Molecular 3D geometry generation; long-range interactions; SE(3)-equivariance.
[PDF]    [Code]
MARS: Markov Molecular Sampling for Multi-objective Drug Discovery
Yutong Xie,  Chence Shi,  Hao Zhou,  Yuwei Yang,  Weinan Zhang,  Yong Yu,  Lei Li 
9th International Conference on Learning Representations (ICLR 2021)  
Multi-objective molecular graph generation via MCMC.
Spotlight Presentation
[PDF]    [Code]
A Graph to Graphs Framework for Retrosynthesis Prediction
Chence Shi,  Minkai Xu,  Hongyu Guo,  Ming Zhang,  Jian Tang 
37th International Conference on Machine Learning (ICML 2020)  
Retrosynthesis via graph translation; inspired by the disconnection approach in Organic Synthesis.
[PDF]    [Code]    [Video Recording]    [Slides]
GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation
Chence Shi*,  Minkai Xu*,  Zhaocheng Zhu,  Weinan Zhang,  Ming Zhang,  Jian Tang 
8th International Conference on Learning Representations (ICLR 2020)  
Molecular graph generation and optimization; flow-based generative models.
[PDF]    [Code]    [Video Recording]    [Slides]
AutoInt: Automatic Feature Interaction Learning via Self-Attentive Neural Networks
Weiping Song,  Chence Shi,  Zhiping Xiao,  Zhijian Duan,  Yewen Xu,  Ming Zhang,  Jian Tang 
28th International Conference on Information and Knowledge Management (CIKM 2019)  
Click-through Rate Prediction with transformers.
A well-known method in the industry
[PDF]    [Code]    [Slides]

(* equal contribution)

Open-source Library

• TorchDrug: A powerful and flexible machine learning platform for drug discovery.
[Homepage]    [Github]    [Twitter]    [Google Colab]

• Recommender Systems: Code base on different recommendation topics, a comprehensive reading list and a set of bechmark data sets.
[Github]

Professional Services

Program Committee member / Reviewer: ICML'21-23, NeurIPS'21-23, ICLR'22-23



Updated at Dec. 2024