Shuxin Zheng
About
Dr. Shuxin Zheng graduated from the MSRA-USTC Joint PhD Program. He currently serves as the Deputy Director of Zhongguancun Institute of Artificial Intelligence, Director of the Strategic Planning Office at the Beijing Zhongguancun Academy, Enterprise Mentor at Chinese Academy of Sciences, and Associate Editor of "AI for Science" journal. Previously, he was a Principal Researcher at Microsoft Research and Head of Microsoft Scientific Foundation Models. He has won multiple world championships in artificial intelligence and trained the largest scientific foundation model to date. Dr. Zheng has published over 20 first-author and corresponding-author papers in Nature subsidiary journals and top international AI conferences, with more than 5,000 citations. He serves as a visiting lecturer at Tsinghua University, Chinese Academy of Sciences, and Microsoft AI Academy, where he has long taught courses such as "Fundamentals of Machine Learning Methods and Applications" and "Advanced Machine Learning."
Research Activities
Education
PhD in Computer Science
University of Science and Technology of China (USTC) & Microsoft Research Asia
BSc in Computer Science
University of Science and Technology of China (USTC)
Teaching Experience
Advanced Machine Learning
graduate courseInstitute of Computing Technology, Chinese Academy of SciencesSpring 2021
Foundations of Machine Learning
undergraduate general education courseTsinghua UniversityFall 2021
Advanced Machine Learning
graduate courseDepartment of EE, Tsinghua UniversitySpring 2021
Advanced Machine Learning
in AI SchoolMicrosoftFall 2020
Advanced Machine Learning
graduate courseDepartment of EE, Tsinghua UniversityFall 2019
Publications
Predicting equilibrium distributions for molecular systems with deep learning
Nature Machine Intelligence2024140 citations
Do Transformers Really Perform Badly for Graph Representation?
Thirty-Fifth Conference on Neural Information Processing Systems (NIPS)20211760 citations
On layer normalization in the transformer architecture
Proceedings of the 37th International Conference on Machine Learning20201360 citations
Asynchronous stochastic gradient descent with delay compensation
Proceedings of the 34th International Conference on Machine Learning, PMLR2017406 citations
Invertible Image Rescaling
European Conference on Computer Vision (ECCV) 20202020309 citations
Cross-Iteration Batch Normalization
IEEE Conference on Computer Vision and Pattern Recognition (CVPR)2020154 citations
One transformer can understand both 2d & 3d molecular data
ICLR 20222022134 citations
Deep learning for prediction of the air quality response to emission changes
Environmental science & technology2020116 citations
Benchmarking graphormer on large-scale molecular modeling datasets
arXiv preprint202288 citations
How could Neural Networks understand Programs?
Proceedings of International Conference on Machine Learning (ICML)202175 citations
Molecule generation for target protein binding with structural motifs
The eleventh international conference on learning representations202372 citations
The impact of large language models on scientific discovery: a preliminary study using gpt-4
arXiv preprint202370 citations
Your transformer may not be as powerful as you expect
Advances in Neural Information Processing Systems202265 citations
Stable, Fast and Accurate: Kernelized Attention with Relative Positional Encoding
Advances in Neural Information Processing Systems, 2021 (NeurIPS 2021)202160 citations
Scalable emulation of protein equilibrium ensembles with generative deep learning
Science202541 citations
π-SGD: Optimizing ReLU Neural Networks in its Positively Scale-Invariant Space
Proceedings of the 7th International Conference on Learning Representations201840 citations
Invertible rescaling network and its extensions
International Journal of Computer Vision202336 citations
Quantized training of gradient boosting decision trees
Advances in neural information processing systems202233 citations
Modeling Lost Information in Lossy Image Compression
arXiv preprint202032 citations
Capacity control of relu neural networks by basis-path norm
Proceedings of the 33rd AAAI Conference on Artificial Intelligence201827 citations