“学萃讲坛”第880期—深度神经网络与主题建模

时间:2018-11-22作者:文章来源:计算机科学与技术学院浏览:36

    “学萃讲坛”秉承学名家风范、萃科技精华的理念,以学术为魂,以育人为本,追求技术创新,提升学术品位,营造浓郁学术氛围,共品科技饕餮盛宴!
报告题目:深度神经网络与主题建模
报   告 人:Wray Buntine
报告时间:2018年11月21日  9:00
报告地点:21#426多媒体报告厅
主办单位:科学技术研究院
承办单位:计算机科学与技术学院
报告人简介:Wray  Buntine is a full professor at Monash University from 2014 and is director of  the Master of Data Science, the Faculty of IT's newest and in-demand degree. He  was previously at NICTA Canberra, Helsinki Institute for Information Technology  where he ran a semantic search project, NASA Ames Research Center, University of  California, Berkeley, and Google. He is known for his theoretical and applied  work and in probabilistic methods for document and text analysis, social  networks, data mining and machine learning.
报告内容:Something Old: In this talk  I will first describe some of our recent work with hierarchical probabilistic  models that are not deep neural networks. Nevertheless, these are currently  among the state of the art in classification and in topic modelling:  k-dependence
Bayesian networks and hierarchical topic models, respectively,  and both are deep models in a different sense. These represent some of the  leading edge machine learning technology prior to the advent of deep neural  networks. Something New: On deep neural networks, I will describe as a point of  comparison some of the state of the art applications I am familiar with:  multi-task learning, document classification, and learning to learn. These build  on the RNNs widely used in semi-structured learning. The old and the new are  remarkably different. So what are the new capabilities deep neural networks have  yielded? Do we even need the old technology? What can we do next? Something  Borrowed: to complete the story, I'll introduce some efforts to combine the two  approaches, borrowing from earlier work in statistics.