Founded in April 2018 by Dr. Junchi Yan, the vision of SJTU-ThinkLab (上海交通大学思维工场实验室) is to develop cutting-edge technolgies for solving the real-world problems whereby data-driven approach interleaved with domain knowledge can be well explored and applied. In particular, the recent years' research is focused on relation & sequence modeling and leanring, which is believed is the main aspects of data science and machine learning.
My recent research is focused on un/semi-structured relational and sequential data analytics and learning, especially for graphs and images (embedding and matching) and temporal sequence (point process and RNN models). I am currently an research professor (PI) with Department of Computer Science and Engineering, Shanghai Jiao Tong University and also the founder of the ThinkLab in SJTU. I am also the member of SJTU Artificial Intelligence Institute and teaches at SJTU Zhiyuan College.
Before joining SJTU, I technically led a team on cognitive IoT for delivering IBM's analytics product and solution based on image and other sensor data with IBM Research - China. I am also the founder of the SDALA lab affiliated with the institute of AIDA@ECNU, whose members are graduate students from ECNU, SJTU and THU when I was with ECNU as an adjunct post-doc researcher during 2015-2018. I coadvised three master students in ECNU and help them publish 2 AAAI paper and 1 ICMR paper in two years, all are first-authored by students.
I authored 20+ filed US patents and first-authored 10+ papers and coauthored 20+ papers in ICCV/CVPR/ECCV/AAAI/IJCAI/TPAMI/TIP/TCYB in the past five years. I serve as an Associate Editor for IEEE ACCESS (IF=3.557, 2017). I am also an adjunct graduate advisor affiliated with School of Data Science of Fudan University.
I have some work on graph matching, optical flow estimation and event data modeling. My current effort is devoted to unifying discrminative and generative model for event sequence modeling -- an important problem in sequential data analytics. I am also interested in applications where sequential data is a big play, such as literature citation, clinical trials, failure logs, social media etc.