时间: 2017年4月17日（周一）上午10:00

地点: 校本部翔英大楼808室

讲座: 半监督学习的近期进展

演讲者: 宫辰教授，南京理工大学

讲座摘要：

Semi-supervised learning (SSL) has been studied for a long history, which aims to classify a massive number of unlabeled examples given the existence of only a few labeled examples. In this talk, I will briefly review our recent works on graph-based SSL, which include three algorithms such as “Fick’s Law Assisted Propagation” (FLAP), “Label Prediction via Deformed Graph Laplacian” (LPDGL), and “Teaching-to-Learn and Learning-to-Teach” (TLLT). Specifically, FLAP utilizes a well-known physical rule called “Fick’s Law of Diffusion” to model the label propagation on a graph, LPDGL incorporates a novel local regularizer to handle the noisy data points, and TLLT builds an interactive machine learning and machine teaching framework to enhance the performance of SSL. The mathematical models, theoretical results, and various practical applications regarding these three algorithms will be discussed in this talk.

演讲者简介：

宫辰于2010年获得华东理工大学学士学位，并于2016年获上海交通大学（SJTU）和悉尼科技大学（UTS）双博士学位。入选南京理工大学“青年拔尖人才选聘项目”，现任南京理工大学计算机科学与工程学院教授。已在国际顶级期刊或会议上发表30余篇学术论文，包括IEEE T-NNLS、IEEE T-IP、IEEE T-CYB、CVPR、 AAAI、IJCAI等。另外，共有5项发明专利获得授权，其中3项为第一完成人。曾获国家奖学金、中国留学基金委“优秀自费留学生”奖学金、IBM优秀学生奖学金、上海交通大学优秀博士学位论文等荣誉。目前担任IEEE T-NNLS、IEEE T-IP、IEEE T-KDE、IEEE T-MM、IEEE T-ITS等19家国际权威期刊审稿人，并受邀担任IJCAI、AAAI、ACPR、ACML等多个国际会议的PC member。