The Division of Science and Technology (DST) held a lecture on “An Unbiased Stratified Statistic and a Fast Gradient Optimization Algorithm Based on It” given by guest speaker Prof. Aixiang Chen on October 22, 2021. Prof. Chen introduced a novel unbiased stratified statistic  and a new algorithm named MSSG based on . He then made comparison with other existing methods.

Prof. Chen giving a lecture 

Prof. Chen is the Associate Dean of Guangdong University of Finance & Economics, a visiting scholar at Wayne State University and Peking University, and the author of the frontier monograph on artificial intelligence titled by Deep Learning. He has presided over 5 scientific research projects, including a national social science project.

Prof. Chen explained the reason for proposing this new statistics  and the construction procedures. He then discussed the variance of unbiased statistic and introduced a novel algorithm MSSG based on this statistic.

Prof. Chen applied the MSSG algorithm to practice and showed the comparison results with the other three algorithms: Batch, SGD and , and concluded that the training and the test accuracy of are the best in most conditions.

Prof. Chen telling the comparison results 

At the end of the lecture, Prof. Chen interacted with the audience and answered their questions.

Prof. Chen answering questions

Prof. Jianxin Pan (right) presenting UIC souvenirs 

Prof. Chen and UIC teachers

Reporter: Xinhui Liu

Photographer: Xinhui Liu

Editors: Ms. Garbo Hu & Ms. Guoqiu Zhang