报告人：钱超
报告地点：腾讯会议560-900-361
报告华体会网页版登陆入口：2022年10月26日星期三下午3:00
报告题目：Multi-objective Evolutionary Learning: Advances in Theories and Algorithms
报告简介：
Machine learning tasks often involve multiple objective functions, which may make conventional optimization algorithms fail. Evolutionary algorithms (EAs), inspired by natural evolution, have yielded encouraging outcomes for solving multi-objective optimization problems in machine learning. However, due to the heuristic nature of EAs, most outcomes to date have been empirical and lack theoretical support. In this talk, I will introduce our efforts towards building the theoretical foundation of multi-objective evolutionary learning. First, I will introduce a general theoretical tool for analyzing multi-objective EAs. Based on this tool, I will then present the influence of recombination operators on the performance of multi-objective EAs, and the theoretical findings on how to deal with constraints and noise. Finally, I will introduce multi-objective evolutionary learning algorithms with provable approximation guarantees, inspired by the theoretical results, for two representative learning tasks, selective ensemble and subset selection.
报告人简介：
钱超，南京大学人工智能学院副教授、博导。分别于2009年和2015年获南京大学计算机系学士和博士学位，之后加入中科大注册登录担任副研究员，2019年回到母校工作。目前主要研究演化计算与演化学习，出版专著《Evolutionary Learning》，并以第一/通讯作者在国际一流期刊和会议（AIJ、TEvC、ECJ、Algorithmica、TCS、AAAI、IJCAI、ICLR、NeurIPS）上发表三十余篇论文，部分成果成功应用于华为工厂排产、无线网络优化等项目，落地华为生产系统。担任IEEE计算智能学会Theoretical Foundations of Bio-inspired Computation工作组主席、IEEE演化计算技术委员会委员、IEEE Trans. on Evolutionary Computation副编、Theoretical Computer Science客座编辑，在国际人工智能联合大会IJCAI’22作Early Career Spotlight报告。获ACM GECCO’11最佳理论论文奖、IDEAL’16 最佳论文奖、IEEE CEC’21最佳学生论文奖提名，博士论文获中国人工智能学会优博，并入选中国科协青年人才托举工程（2016），获国家优秀青年科学基金（2020）。
