邮 件：jwsu@ustc.edu.cn
所属单位：人工智能与数据科学学院
个人主页：https://jwsu825.github.io/
主要研究方向：机器学习算法–系统协同设计，学习理论与优化算法，AI for Science，图网络学习
个人简介
苏骏炜，预聘副教授，中国科学技术大学人工智能与数据科学学院。主要从事机器学习理论、算法与系统研究，聚焦高效可信机器学习的算法–系统协同设计。博士期间获香港博士研究生奖学金（HKPFS）和香港大学校长奖学金资助，长期担任 NeurIPS、ICML、ICLR 等国际顶级会议的 Area Chair 和审稿人。本科毕业于英属哥伦比亚大学，硕士毕业于多伦多大学，博士毕业于香港大学。长期围绕图学习与网络建模、大规模机器学习算法与系统、学习理论等方向开展研究，在 ICML、ICLR、KDD、VLDB、EuroSys、AAAI、ECCV、SIMODS 等国际顶级会议和期刊发表论文二十余篇。
教授课程：暂无
近几年代表论著
[1] Su Junwei，Zou Difan，Zhang Zijun，Wu Chuan. Towards Robust Graph Incremental Learning on Evolving Graphs.Proceedings of the International Conference on Machine Learning (ICML), 2023.
[2] Su Junwei，Zou Difan，Wu Chuan. PRES: Toward Scalable Memory-Based Dynamic Graph NeuralNetworks. Proceedings of the International Conference on LearningRepresentations (ICLR), 2024.
[3] Sheng Guangming, Su Junwei(Corresponding), Huang Chao, WuChuan. Mspipe: Efficient temporal gnn training via staleness-aware pipeline.Proceedings of the Conference on Knowledge Discovery and Data Mining (KDD),2024.
[4] Hu Hanpeng, Su Junwei(Corresponding), Zhao Juntao, PengYanghua, Zhu Yibo, Lin Haibin, Wu Chuan. CDMPP: A device-model agnosticframework for latency prediction of tensor programs. Proceedings ofthe European Conference on Computer Systems (EuroSys), 2024.
[5] Su Junwei, WuChuan. On the topology awareness and generalization performance of graph neuralnetworks. Proceedingsof EuropeanConference on Computer Vision (ECCV), 2024, best papercandidate.
[7] Su Junwei, WuShan. Temporal-Aware Evaluation and Learning for Temporal Graph NeuralNetworks. Proceedingsof the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI), 2025.
[8] Zhong Yuchen,Su Junwei(Corresponding), Wu Chuan, Wang Minjie. Heta: Distributed Training ofHeterogeneous Graph Neural Networks. Proceedings of the Very Large Data Base Conference (VLDB),2025.
[9] Su Junwei, Wu Chuan.On the Interplay between Graph Structure and Learning Algorithms in GraphNeural Networks. Proceedingsof the International Conference on Machine Learning (ICML),2025.
[10] Su Junwei，Wu Shan. SBGD: Improving Graph Diffusion Generative Model via StochasticBlock Diffusion. Proceedingsof the International Conference on Machine Learning (ICML),2025.
[11] Su Junwei，Wu Chuan. A Non-Asymptotic Convergent Analysis for Scored-Based GraphGenerative Model via a System of Stochastic Differential Equations. Proceedings of theInternational Conference on Machine Learning (ICML),2025.
[12]Qi Guicheng，Su Junwei(Corresponding)，Yang Liqi，LiTao，XieTingwen，Sun Yerui，XieYuchen，Wu Chuan. HetAuto: Cross-Cluster Auto-Parallelism forHeterogeneous Distributed Training.Proceedings of the European Conference on ComputerSystems (EuroSys), 2026.
[13]Liu Mengfan, Zheng Da, Su Junwei (Corresponding), Wu Chuan. Full-Graph vs. Mini-Batch Training:Comprehensive Analysis from a Batch Size and Fan-Out Size Perspective. Proceedings of the International Conference on Learning Representations(ICLR), 2026.
[14] Su Junwei,Marbach Peter. The Role of Social Support and Influencers in Content Markets.ACM Transactions on Economics and Computation.
[15]SuJunwei, Wu Chuan, Zheng Le , Zou Difan. Improving Implicit Regularization ofSGD with Preconditioning for Least Square Problems. SIAM Journal on Mathematicsof Data Science