加入收藏夹
联系我们
关于本站
个人主页
西电导航
西安电子科技大学
    当前位置:首页>>学术讲座
Sparse and Deep Multi-Task Learning with Applications to Health Informatics
时间:2018-06-23 22:46    点击:   所属单位:通信工程学院
讲座名称 Sparse and Deep Multi-Task Learning with Applications to Health Informatics
讲座时间 2018-06-27 10:00:00
讲座地点 北校区新科技楼1012会议室
讲座人 Prof. Dongxiao Zhu
讲座人介绍 Dongxiao Zhu is currently an Associate Professor at Department of Computer Science, Wayne State University. He also directs the Data Science and Machine Learning research lab. He received his Ph.D. from University of Michigan in 2006, Master from Peking University and Bachelor from Shandong University. His research interests have been in areas of machine learning and data science with applications to big data in bioinformatics, health informatics, multimedia and natural language processing. Dr. Zhu has published over 60 peer-reviewed publications and numerous book chapters and he served on several editorial boards of high-impact scientific journals. Two of Dr. Zhu's recent machine learning papers won the "Best Paper Award Top 3 Finalist" and "Best Poster Award Top 3 Finalist" at IEEE-ICMLA 2017. Dr. Zhu's research has been supported by National Institute of Health (NIH), National Science Foundation (NSF) and private agencies and he has served on multiple NIH and NSF grant review panels. Dr. Zhu has advised numerous students at undergraduate, graduate and postdoctoral levels and his teaching interest lies in programming, data structures and algorithms, machine learning and data science
讲座内容 Many machine learning problems are inherently multi-task where tasks can represent multiple subgroups of input or multiple labels. Building an individual predictive model for each task can lead to overfitting due to limited labelled samples and high input dimension. Here we develop a generalized multi-task learning framework that bridges data from all tasks and improves their generalization performance. Our approach is sufficiently flexible that leverages both deep neural networks and a variety of regularization terms for an efficient and effective feature selection. We demonstrate the superior performance of our approaches in three scenarios, i.e., multi-task survival analysis, multi-task ordinal regression and multi-task learning with auxiliary tasks.     
转载请注明出处:西安电子科技大学学术信息网
如果您有学术信息或学术动态,欢迎投稿。我们将在第一时间确认并收录,投稿邮箱: meeting@xidian.edu.cn
Copyright © 2011-2018 西安电子科技大学 
开发维护:电子工程学院网络信息中心  管理员:meeting@xidian.edu.cn 站长统计: