弱监督学习
Weakly supervised learning, a probabilistic perspective
刘同亮   Liu Tongliang
报告人照片   Tongliang Liu received the BEng degree in EEIS from the University of Science and Technology of China and the PhD degree in IS from the University of Technology Sydney. He is currently a Lecturer with the School of Computer Science and the Faculty of Engineering, and a core member in the UBTECH Sydney AI Centre, at The University of Sydney. His research interests include statistical learning theory, machine learning, computer vision. He has authored and co-authored 60+ research papers including IEEE T-PAMI, T-NNLS, T-IP, ICML, CVPR, ECCV, AAAI, IJCAI, and KDD. He is a recipient of Discovery Early Career Researcher Award (DECRA) from Australian Research Council (ARC) and was shortlisted for the J G Russell Award by Australian Academy of Science (AAS) in 2019.
  Most machine learning algorithms rely heavily on accurate supervisory information. However, as datasets grow bigger and bigger, obtaining accurate labels becomes laborious, time-consuming, and expensive. Weakly supervised data is therefore becoming popular. In this talk, we show that for many weakly supervised settings, such as semi-supervised learning, positive and unlabelled (PU) learning, multi-instance learning, and label noise learning, the classifier learned from the weakly supervised data could converge to the optimal classifier learned by employing accurately labelled data. We discuss challengings and solutions in designing such consistent learning algorithms.
报告时间:2019年08月11日15时30分    报告地点:西区科技实验楼西楼118会议室
报名截止日期:2019年08月11日    可选人数:60