BIMK
Institute of
Bioinspired Intelligence and Mining Knowledge
Location:[an error occurred while processing this directive]

金耀初教授学术报告会

2018-04-28

Recent Advances in Infill Criteriafor Surrogate-AssistedEvolutionaryOptimization

 

Presenter: Professor Yaochu Jin

Department of Computer Science, Universityof Surrey

Email: yaochu.jin@surrey.ac.uk

URL: http://www.surrey.ac.uk/cs/research/nice/people/yaochu_jin/

 

报告地点:安徽大学理工D楼318

报告时间:2018年5月2日上午9:30-10.30

 

Abstract:

Infill criteria have been shown to be effective formodel management in Gaussian process assisted evolutionary optimization oflow-dimensional expensive problems. However, they are prevented from beingapplied to high-dimensional problems due to two serious problems. First, thecomputational complexity of constructing Gaussian processes increasesdramatically when the number of training samples increases. Second, theuncertainty information estimated by the Gaussian process becomes lessreliable, degrading the effectiveness of the infill criteria. This talk presents a brief account of recent advances in addressing the above challenges in  surrogate-assisted evolutionary optimization of high-dimensional problems. Heterogeneous ensemble learningmodels are generated to replace the Gaussian process to improve the scalabilityof computational complexity for high-dimensional problems. In addition, amulti-objective infill criterion has been proposed to replace existing singleobjective infill criteria. Empirical results demonstrate the effectiveness ofthe proposed methods.

 

SpeakerBio:

金耀初分别于1988、1991及1996年在浙江大学电机系获学士、硕士及博士学位,并于2001年在德国波鸿鲁尔大学神经信息研究所获工学博士学位。现为英国萨里大学计算科学系“计算智能”首席教授, “自然计算与应用”研究组主任,萨里大学“数学与计算生物学中心”共同负责人。金耀初博士是中组部“千人计划”专家,教育部“长江学者”讲座教授,芬兰国家技术创新局“芬兰讲座教授”。目前担任《IEEE认知与发育系统汇刊》主编, Springer《复杂与智能系统》共同主编,IEEE 杰出演讲人。IEEE Fellow。已出版专/编著及会议论文集9本,发表学术论文200余篇。论文被引用总次数13500余次 (据GoogleScholar), 其中SCI引用4200余次,h-index 为55。获美国、欧盟和日本专利共9项。先后在30多个国际会议上作特邀大会或主题报告。