讲座题目🦝：Evolutionary Computation: A Path from Biology to Artificial Intelligence 主讲人：Erik D. Goodman 教授 主持人🧖🏽：周爱民 副院长 开始时间🙆🏻♂️：2019-10-29 14:00:00 讲座地址🙏🏻：中北校区数学馆201 主办单位：计算机科学与技术学院
报告人简介🆑🎼： Erik D. Goodman教授是进化计算领域国际著名学者🧑🏭🛂。现任密歇根州立大学电子与计算机工程系教授、美国国家科学技术研究中心“生物进化在行动”BEACIN（Bio/computational Evolution in Action CONsortium Center）的中心主任🤦🏻♂️。他曾担任第一届进化计算及其应用国际会议主席，2001-2004年担任国际遗传进化计算协会创始主席💇🏽♂️，2005-2007年间担任美国计算机学会（ACM）基因及进化计算特别兴趣小组（SIGEVO）的创始人及第一届主席🔫。自2008年起🤼，成为纽约城市大学John Jay学院突发紧急事件分析学院的指导委员会核心成员📬。他曾是同济大学🧖🏼、华东师大🔜、汕头大学等国内多所高校的顾问教授。 Erik D. Goodman教授还获得了大量的奖项🧑🏻🔬，包括2011年的密歇根州立大学的杰出教员奖，2011年ACM SIGEVO 杰出服务奖（以表彰作为ACM SIGEVO 创始人所做的贡献）🚲，2009年的密歇根州杰出教授奖，2007年的密歇根州立大学最高教学奖。他还当选为乌克兰国家工程学院院士🧓🏻𓀃、及国际信息科学院院士🚉。担任包括进化计算领域权威期刊《Evolutionary Computation》等在内的多个国际期刊编委。 报告内容🏄🏽♀️： In its early days, prior to about 1990, evolutionary computation (genetic algorithms, genetic programming, evolution strategies, etc.) used only a very simple model of evolution in nature, in part because computers were too slow to handle difficult real-world problems that required more complex algorithms to solve. But researchers began to take more lessons from nature, resulting in more powerful evolutionary algorithms that capture the effects of distinct subpopulations experiencing different selective pressures, resulting in maintenance of far more diversity in the overall population. Collaborative problem solving by human teams has also informed the structure of newer evolutionary algorithms, yielding the ability to solve much more difficult real-world problems, as examples to be presented will demonstrate. But these strong capabilities were largely ignored by the machine learning (ML) community until deep convolutional neural networks became popular. Evolutionary algorithms have shown that they can efficiently evolve improved architectures for DNNs, which is most important when the resulting nets are to be deployed in cell phones and other devices with strong hardware limitations. The big Deep Learning companies are now hiring evolutionary computation experts to complement their ML strength. |
