Professor Zhi-Hua Zhou, Nanjing University, gave a talk entitled " Evolutionary Learning: Advances in Theories and Algorithms" at the School of Computer and Electronic Information on Dec. 3, 2021. The academic presentation was hosted by Professor Genlin Ji, Dean of the School of Computer and Electronic Information.

 

Zhi-Hua Zhou, Academician of the European Academy of Sciences, Chairman of the Jiangsu Computer Society, Chairman of the Jiangsu Artificial Intelligence Society, Professor of Nanjing University, Member of the Academic Committee, Dean of the Computer Department, and Dean of the School of Artificial Intelligence, mainly engaged in machine learning and artificial intelligence research. He has made important contributions in ensemble learning, multi-label learning and weakly supervised learning. He is the author of "Machine Learning," "Ensemble Methods: Foundations and Algorithms," etc., and his works have been cited more than 60,000 times. His research results have been transformed and implemented in enterprises such as Huawei and major national projects. As the first complete person, he has won the National Natural Science Second Prize twice, the Ministry of Education Natural Science First Prize three times, and the IEEE Computer Society Edward J. McCluskey Technical Achievement Award, CCF Wang Selection Award, etc. He is ACM, AAAI, IEEE, etc. Fellow. He has served as a consultant for AI Magazine and chairperson of the program committee of the 2019 International Artificial Intelligence Conference and the 2021 International Artificial Intelligence Joint Conference.

Professor Zhou first introduced the basic problems of evolutionary learning, the development process and multi-objective optimization problems in machine learning. He then introduced the main analysis methods and important theoretical results of the evolutionary learning theoretical problems and gave a general method for boundary and characterization of approximation performance. Finally, Professor Zhou shared the theoretically guaranteed and excellent performance selective integration algorithm proposed by his team and proposed an evolutionary learning algorithm with better performance than the classic optimal algorithm for the "noisy subset selection" problem.